Power and Ownership Structures among German Companies∗ A Network Analysis of Financial Linkages
Jochen Moebert and Patrick Tydecks
June 21, 2007
Abstract
The literature on ownership structures has made continual use of notions such as cross-holdings and pyramids which are closely related to the vastly increasing network literature. We propose to transfer successfully applied network methods such as network graphs, the MAN-classification scheme, and centrality concepts to the corporate control and corporate governance branch as well. Given these concepts and a unique data set containing 2784 companies we can identify the most powerful German companies and their characteristics.
JEL Classification: G32, L14
Keywords: network, ownership structure, corporate control, power, financial linkages
Contact details of the authors:
Darmstadt University of Technology, Institute of Economics, Applied Economic Research and Microeconometrics, Marktplatz 15, Residenzschloss, D-64283 Darmstadt, Germany, [email protected] (Jochen Moebert) [email protected] (Patrick Tydecks)
∗For comments we are grateful to Horst Entorf, Tobias Klein, Alexander Ludwig, Martin Salm, and all other seminar participants at the University of Mannheim. We are also grateful to Ulrich M¨uller, Christoph Wachtel, Alexander G¨orbing, and Thorsten Jacoby of Hoppenstedt who provide the data base. Igor Goncharev from the ‘Monopolkommission’ explained us the way indirect relationships are calculated in the ‘Hauptgutachten’ published in biennial updates. We gratefully acknowledge the assistance of Philip Savage for the rectification of our first version. Jochen M¨obert extends his thanks to Martin Hellwig, Christian Laux, and Holger M¨uller who introduced him into this topic during seminars and lectures.
1 I Introduction
Germany’s corporate control system has at least three dimensions, i.e. the supervisory board, voting control at general assemblies (cf. Becht and Boehmer (2003)), and ownership stakes. 1
All three are closely related, but due to several specialities in German corporate law the impact of each dimension on the power of a company can be different. Due to ownership structures and these specialities cash flows and voting power are separated in many cases (cf. Bebchuk et al. (2000), among others). In this work we identify the most powerful companies with respect to ownership stakes. The identification is important since, in Germany in particular, powerful companies might be large shareholders and such blockholders can have both a beneficial and detrimental influence. 2 Large shareholders can misuse their power in takeover proposals. A blockholder of both the absorbing company and the acquiring company profits if a firm is sold below its value whereas minority shareholders of the acquired company are exploited. Such blockholder strategies which expropriate minority shareholders are often described by the term
‘tunneling’ in the corporate governance literature (cf. Bertrand et al. (2002)). Franks and
Mayer (2001) analysed tunneling effects among German companies in 1990 and found little evidence that tunneling was an important issue. However, the misuse of tunneling effects are documented in several other cases. Attig et al. (2003) compile a data set on Canadian stock corporations and find evidence for the misuse of power which the ultimate owners of pyramids hold. They argue that ultimate owners maximize profits at the expense of minority shareholders and companies which have a high distance to the ultimate owners. Meoli et al. (2006) elaborate the Telecom Italia case where minority shareholders are expropriated by specific network and dual-class structures and Atanasov (2005) documents the malpractice of tunneling during the mass privatization in Bulgaria. There is also some evidence that tunneling effects played a role during the financial downturn in the Asian crisis (cf. Johnson et al. (2000), Lemmon and
Lins (2003)). 3 Hence, given these cases and the globalised financial markets detrimental consequences of block ownership can be seen as an issue in Germany as well.
In this article we exploit the large data sources available to get a micro-picture of a large part of the German corporate landscape which provides the basis for thorough investigations
2 in the future. To put into practice such a micro-macro perspective a network approach is a rational choice. This methodology offers both a local micro-perspective since each company and its shareholders can be analysed alone as well as offering a bird’s-eye macro-view due to the interconnectedness of all firms. Furthermore and particularly, the interaction between the micro and macro level can be investigated. Network science offers a variety of ideal tools for the development of real micro-based macroeconomics. The network perspective was already implicitly postulated in the literature on corporate control. La Porta et al. (1999) argued
“For most countries, this [network perspective 4] is the only way to understand the relationship between ownership and control” 5 (cf. also Faccio and Lang (2002), Chapelle and Szafarz (2005), among others). Hence, we analyse a large network data set and thereby extend the literature on company networks. Our aim is the identification of the most powerful companies. We measure the power of companies by centrality concepts. For decades these statistics have been successfully applied in the social network literature. Therefore, our work is based on the company network literature and the vast corporate governance literature.
In addition, to the identification of powerful companies our analysis shed some light on the
‘Deutschland AG’. Due to its particularities, the German case is intensively debated in the corporate governance literature. However, many statements are based on a small data base which concentrates especially on large companies such as listed stock corporations or the largest one hundred companies. 6 The German ownership structure called ‘Deutschland AG’ was (or still is) so isolated relative to Anglo-Saxon markets and interwoven that corporate control was implicitly exerted by the national companies themselves. Hence, legitimate ownership rights were disregarded and corporate control from outsiders, such as international shareholders as well as other stakeholders, was limited. This corporate network restrained non-national firms from gaining a foothold in the German company system and specific ownership structures among major companies hindered hostile takeovers.
In recent years, it has been much discussed that the corporate ownership structure is subject to change in Germany. Due to the globalization and tax abatements on capital gains realized by sale, many blockholders diversified their investment portfolio by adding international companies
3 and cutting down national holdings. In particular, bank and insurance companies have changed their investment portfolios. Therefore, long-term relationships which often existed for decades, especially between banks and industrial companies, were broken. 7 These often mentioned breakups of bank-industry links being formed during the period of industrialisation (cf. Franks et al. (2005) and Fohlin (2005) who provide us with insights into the historical development of
financial linkages in Germany), are the origin of statements such as “Is Deutschland AG kaputt?”
8 However, such analyses define the ‘Deutschland AG’ as a sub sample of all important German companies whereas all important ones might be of interest. Hence, the dissolution process observed may be nonexistent in the basic population. Our data set instead contains many more companies than other studies and may be more informative. Furthermore, the dissolution process observed among large national firms might be replaced by linkages to international connections. Becht and R¨oell(1999) have documented that companies from many countries in continental Europe have large voting blocks. Possibly, the dissolution among national firms is being replaced by new financial linkages among European corporations. 9
Before we start our network analysis, we will review important contributions to the company network literature in Section I. Subsequently, in Section II we describe our data set. In Section
III the most central corporations are identified by the application of standard network concepts.
Furthermore, in a subsection to Section III firm characteristics are used to explain the centrality vector in an econometric model. Hence, we can identify the industries and the firm characteristics which are related to a high or low centrality and Section IV concludes. For many readers the
Appendix might also be of interest. Many network structures of large German companies are shown there.
II Company Network Literature
The network literature on the ‘Deutschland AG’ and financial interlocking of firms is neither very detailed nor exhaustive. However, there are some important contributions which are first steps towards a deeper understandings of corporate ownership structures. These articles mainly
4 written by social scientists are briefly reviewed here. To zero in on important contributions we review papers focused on the German company network. Moreover, we casually include contributions concerned with firm networks from other countries or with interlocking directorates in Germany. 10
A large network study with respect to size was performed by Kogut and Walker (2001), who used data from the Frankfurter Allgemeine Zeitung GmbH. 11 They investigated how the German ownership network influences merger and acquisition activities from 1993 to 1997. Their firm sample incorporates the largest five hundred non-financial companies, the 25 largest banks, and the 25 largest insurers in 1993. After the selection of this sample the 684 owners of these
550 firms were ascertained. Finally, a binary network of zeros and ones among companies was arranged. The ones represent all direct links if the equity stake of a shareholder was above
5 percent. Hence, this network formation process ignores all blockholders below 5 percent and equally weighted all stakes above 5 percent. They therefore ignore a large part of small shareholders as shown in the next section, where our network data set is analysed. The M&A data base includes 101 acquisitions which take place among the 550 companies from 1994 to
1997. By means of simulation the authors showed that randomly rewiring company holdings affects the German corporate system only slightly. 12 If, for instance, one hundred links are rewired then the average path length only dropped about 20% and the cluster coefficients about
30%. These findings are in line with small world networks and indicate the intrinsic stability of the corporation network. Furthermore, it is argued that mergers and acquisitions maintained the structure of the German company network since very central companies seem to be more active in acquiring firms than the average company in the sample. Therefore, their findings challenge the thesis on the dissolution of the ‘Deutschland AG’.
Heinze (2004) investigated the change of interlocking directorates instead of the financial interlocking of the ‘Deutschland AG’ from 1989 to 2001. He described the different control structures by means of descriptive network statistics and also concentrated on large German companies. Furthermore, he asserts that both the financial network and the personal network of executive and supervisory board members are tightly knit and both networks co-evolved
5 historically. In the twelve year span, many links in the network structure were diluted. But many local network structures such as cliques and core-periphery structures were unaffected.
Furthermore, the financial companies are still the most central players. We are not convinced that financial and personal networks co-evolve similarly. Of course, shareholders can affect board elections. However, German laws establish special rules affecting board composition which dilute the power of shareholders. From a theoretical point of view, both types of links can be seen as substitutes of a common goal national companies share. While globalization and German tax policy boost incentives to abolish equity stakes, members of executive and control boards might be willing to strengthen the ‘Deutschland AG’ by maintaining or intensifying personal relationships.
H¨opner and Krempel (2004) visualized the German company network for 1996 and 2000. The data base includes the one hundred largest companies and is provided by the German Monopolies
Commission, 13 which publishes an official report about the competitive position of German corporations every second year. Inspection by eye reveals that the network density shrinks because several links were severed between financial and industrial companies. In addition, links between financial companies are diluted. As mentioned above, these observations contrast with the stability argument of Kogut and Walker (2001).
In an early study, Pappi et al. (1987) analyse the financial interlocking as well as interlocking directorates of the largest 325 German companies in 1976. The 205 industrial companies were chosen due to the highest turnover level of all companies in 1976. The largest banks are identified by their balance sheet total and the largest insurers were chosen due to a ranking of earned premiums. Each company unit is sectioned into one of ten blocks which are defined by means of a cluster analysis. Subsequently, relationships among the blocks are investigated by analysing personal and financial linkages. Their analysis underpins the power of large German banks in former decades.
Recently, the focus among network researchers turned to the analysis of the historical evolvement of company networks. For instance, Windolf (2005) compares the development of U.S. and
6 German firms between 1896 and 1938. His research suggests that the difference between both countries found today is caused by different developments in the 20th century. Whereas the
financial interlocking is quite similar, the interlocking among members of the supervisory board was much more concentrated in Germany than in the U.S.
III The Data Base
The data base used for the analysis is the Hoppenstedt Konzernstrukturdatenbank 14. The data bank is one of Germany’s major data banks containing ownership structures of more than 250,000 companies. This data source often abbreviated KSD was also used by Becht and B¨ohmer (1997),
Kammerath (1999), K¨oke (1999), Becht and B¨ohmer (2001, 2003), among others. 15 and is also one primary source of the German monopolies commission. 16 The KSD contains self-reported information as well as actively collected information pieces via professional data managers.
The data collection process started on 20th May 2006 and was completed by 20th June 2006.
This process can be separated into four steps. First, we picked all German companies with a turnover of at least one billion euro. This sample includes only single company units but no parent companies which are just holdings or have a turnover below one billion euro. This core sample contains 597 industrial companies. 17 Second, we gathered all direct and indirect ownership relationships among this core sample. Due to definitional issues, the revenues of
financial companies are not termed turnover. The turnover criterion was also chosen by Pappi et al. (1987), however, for the financial companies, we adopted, due to better data availability, a different approach from these authors. The third step was the identification of all direct and indirect parent companies from the first chosen sample of 597 units. These direct and indirect links can be conveniently depicted in Network Figures as shown in the Appendix. Also,
Kogut and Walker (2001) used such network data but took into account only the direct parent companies. We also include the parent companies of the parent companies up to distance six, where the term ‘distance’ in network terminology is defined as the number of links between two companies and offers therefore a much deeper view than earlier work on Germany’s corporate
7 structure. This third step extended the total sample to 2784 companies, which also contained all major German financial companies. Fourth and finally, all shareholder relationships among all firms were compiled. Our network data set is very different from previous work on company structures performed by economists where the focus is mostly on the position of a single company.
For instance, La Porta et al. (1999) provide us with a description of the ownership structure of
Allianz and DaimlerChrysler. 18 This micro-perspective instead of a network view makes the application of network tools unappealing or even impossible.
The close relationship among Allianz, Dresdner Bank, and M¨unchner R¨uck is the classical paradigm of interwoven German companies (cf. La Porta et al. (1999)). In a certain manner, the financial linkages among these three corporations enabled them to bypass German stock corporation law 19 and, correspondingly, hostile takeovers and, more importantly, corporate control of outsiders were virtually impossible even if those firms and executives performed poorly.
Even today, Allianz and M¨unchner R¨uck are important blockholder of each other. The Allianz holds 9.4% of the M¨unchner R¨uck whereas the M¨unchner R¨uck holds 4.9% of all Allianz shares.
Otherwise the Allianz corporation has a dispersed ownership structure. Therefore, the Allianz network consists only of two nodes and two arcs representing the Allianz-M¨unchener R¨uck cross-holding. Due to its simplicity the Allianz network is omitted in the company Network
Figures shown in the Appendix. However, the ego-centered company networks of Aldi, AMB
Generali, AXA, BMW, Commerzbank, DaimlerChrysler, Ergo concern, Metro, Deutsche Post,
Deutsche Telekom, and Volkswagen are depicted in Network Figure 5 to 15 in the Appendix.
The final data set contains industrial and financial companies, state enterprises, partnerships, and individuals. Our data set also offers an international perspective on the German company network since not only national firms but also foreign firms are taken into account. The number of foreign firms amounts to 824 or 29.53% of the sample size. The number of companies from each country relative to the total number of companies in the network is reported in Table I.
Apparently, large economies such as US, UK, Japan, etc. make up the largest number of foreign
firms related to the German company network. Interestingly, firms based in tax havens such
8 as the Cayman Islands and Bermuda have a similarly large number of relationships comparable with companies located in Spain and Canada.
[Insert Table I about here]
Another important firm characteristic is the legal form of companies. Legal forms of different countries are not completely comparable. However, the different types of companies were allocated to different groups in keeping with Table X, as shown in the Appendix. Given this assignment, most companies in our sample are limited companies as documented in Table II.
Expectedly, a large share of private and public limited companies is found. A high number of individuals and state enterprises is also included into the German company network. This
finding is often exposed as one major difference in the shareholder structure of Anglo-Saxon and German companies as well as other companies located in continental Europe. According to Burkart et al. (2003), the large number of family-owned German corporations is caused by weak minority shareholder protection which is often attributed to the poor German corporate governance system. Even after recent changes no stronger market-oriented governance system is assumed (cf. Terberger (2003), Goergen (2004), among others). Hence, the importance of family blockholders will continue to be a feature in the future. Moreover, the number of individuals in our network may underrate their power since individuals and families are often ultimate owners of firms. Faccio and Lang (2002) find that Western European firms are either family controlled or have dispersed ownership structures. Their comparison of ultimate owners across countries unveils the exceptional position of family firms in Germany. For instance, for publicly traded
firms the ultimate owner is a family in about two-thirds of cases and about nine out of ten unlisted German firms are family-owned.
[Insert Table II about here]
For generations shareholders of large German corporations have been well-known families. For instance, the Quandt family holds a large share in BMW and the Pi¨ech family and Porsche family are still among the large blockholders of VW (compare Network Figures 8 and 15). The
9 figures show that these families are not only represented by one company protecting rights of a whole family but that there are quite complex holding structures in which several individuals of each family are involved. Interestingly, individuals are sometimes only indirect blockholders of the automobile corporations since limited companies typically in complete individual ownership lie in between. For instance, Johanna Quandt is the sole owner of Johanna Quandt GmbH &
Co. KG which holds 14.21% of all BMW shares. Often the impact of family ownership on firm performance and corporate control is debated. On the one hand family ownership might facilitate a thorough development of a company, on the other hand block ownership might hinder effective corporate control. Recently, Villalonga and Amit (2006) analysed the impact of family ownership on firm performance and found mixed results for US firms. Nowak et al. (2006) as well as Maury
(2005) report a positive relationship between operating performance and family-ownership.
State enterprises are also involved in many German companies. Again, Network Figure 15 of
VW exemplarily shows a state-firm relationship. The Hannoversche Beteiligungs mbH is a large shareholder of Volkswagen and is owned by the Bundesland Niedersachsen 20. Similarly, the
German state is still engaged in the DAX companies Deutsche Post and Deutsche Telekom imaged in Network Figures 13 and 14. The vast majority of ‘state enterprises’ are owned by medium-sized and large cities which are often connected to public utility companies as well as local saving banks. 21 Interestingly, in their cross-country comparison La Porta et al. (1999) and La Porta et al. (2002) argue that both a relatively high number of family-owned firms and a large influence of government entities indicate insufficient shareholder rights. For Germany, the low degree of shareholder protection relative to Anglo-Saxon countries is often reported and details about German corporate law - briefly discussed in the following paragraph - point out this fact.
III.A Descriptive Network Statistics
Network consists of vertices and arcs between the vertices. In a company network the vertices are the companies themselves and arcs represent the ownership structures among these companies,
10 where the arrows point from the companies to their shareholders. In total, our company network exhibits 3711 arcs and consists of 192 components, where companies of two different network components are neither directly nor indirectly connected. 22 Weights are attached to each arc to capture the different shares being held and the power exerted by owners. However, for the sake of clarity links in network figures shown in the Appendix are categorized into three classes.
The first class summarizes small equity stakes below 10%, the second class contains equity stakes lying in the right open interval from 10% to 50%, and the third class contains equity stakes at or above 50%. In the network figures the three classes have different line widths.
For instance, in Network Figure 15 an arc with a weight of 15.46 goes from Volkswagen to the
Porsche corporation which indicates that Porsche holds 15.46% of all Volkswagen shares. 23
The mode weight in the complete network shown in Network Figure 1 is 100% whereas the mean value is 45.5% and the median is 27.7%. The mode weight is observed in about one third of all links. Obviously, holdings often completely own their subsidiaries. Means and Medians of previous studies are both about 50% - an overview of several other Germany-related studies is documented in Becht and Boehmer (2003) as well as Goergen et al. (2004). Differences between previous studies and our median can be attributed to our larger data base, to different sample periods, or both. Other often observable link weights are equity holdings of about 10%, 20%,
50%, 75%, and about only a few percent as illustrated in Figure 1(a) and (b). Regarding block ownership, our findings are in accordance with previous studies, e.g. La Porta et al. (1998), who found strong concentrations in ownership structure in nearly all countries. Concentrated ownership structure is also induced by Gemany’s Companies Act 24. Germans stock corporation law gives (minority) shareholders specific rights.
For instance, individual discharges of each member of the supervisory board - instead of contemporaneous discharge of all members - is enforceable by shareholders holding 10% of the voting equity (§120(1) AktG - see also §137 AktG). Similarly, an investor requires at least 20% of the voting equity (§122 AktG) to enforce extraordinary general meetings. At least 50% of all votes are necessary to enforce decisions at general assemblies (§153 AktG). Also, the appointment of auditors scrutinising the formation process, the increase of capital, or capital reduction (§142
11 AktG) as well as raising a claim against board members or directors (§147 AktG) explicitly requires an ordinary majority. A qualified interest enables shareholders to amend corporate statues (§ 179 AktG) and to increase in registered capital (§182 AktG). Hence, it is obvious that chosen blockholder stakes are not randomly assigned between firms but are chosen to foster or block specific rules.
Figure 1(c) shows the distribution of incoming arcs (indegree) and the distribution of outgoing arcs (outdegree) of all companies in our network. Both functions are quite similar. The linearity in the log-log diagram indicates that there are a few central companies with many links and many firms who just have a small number of equity stakes. 25 Subfigure 1(d) shows the mean and the difference ∆ of 66 cross-holdings in the total network. 26 Mean and difference are always calculated for each cross-holding. One cross-holding between Allianz and M¨unchner R¨uck was mentioned above and another exists between ‘K¨olnische Verwaltungs-Aktiengesellschaft f¨ur
Versicherungswerte’ and the AXA concern as shown in Network Figure 7. Most cross-holdings such as the Allianz-M¨unchner R¨uck link have capital weights below 10% in both directions, therefore, both the difference and the mean of cross-holdings are small. The second cross-holding in Network Figure 7 has values of 25.631% and 23.02%. Hence, the mean is in the 20% interval whereas the difference lies in the 10% interval in Figure 1(d).
12 III.B MAN-analysis
One powerful mean to analyse the networks is the triad MAN-classification scheme proposed by Holland and Leinhardt (1970). This descriptive statistic is a simple count mechanism which
27 2784 picks all possible combinations of triads among all nodes - in our case there are 3 = 3, 592, 429, 984 triads. After each combination the existing links among the nodes are observed.
There are sixteen possible combinations depicted in Figure 3 in the Appendix representing the
MAN-classification scheme. M represents the number of mutual dyads, A asymmetric dyads, and N null dyads in a triad. In addition, for some triads a letter is added to indicate the direction of the arrows in a triad where D abbreviates down, U up, T transitive, and C cycle.
The MAN-classification scheme measures micro network formations and, contemporaneously, provides access to a macro perspective. All 16 possible triad formations observed in the company network are summarized in Table III. For instance, 003 triads - the triads which contains three null dyads, i.e. no links at all - are found much more often than expected, whereas 012 triads are less often observed than expected. Thereby, the term ‘expected’ refers to a random network where each link has the same probability to be present. Table 8 shows the probabilities to observe certain triad formations in a random network. Our results indicate that the network formation process underlies a non-random process. In total, 003 triads and 012 triads are less often observed than expected. This indicates that certain network formations - those where more than one arc is involved - are likely to emerge. One such triad involving more than one arc contains mutual links, i.e. cross-holdings. The MAN-classification scheme reports a very high relative number of 102 triads as indicated by the ratio of observed to expected triads (O/E ratio in Table III). Hence, we can conclude that firms have a high incentive for cross-holdings.
This micro network structure is often seen as a classical form of ownership concentration. The reciprocal relationship can hinder the exercise of corporate control if the reciprocal voting power is large enough and managers are reluctant to explicitly control each other. Given the relative high number of mutual links it is also not surprising that we observe more 201, 120D, 120U, and
120C triads than expected. However, the absolute number of these triads is fairly low such that these formations are of minor importance.
13 Figure 1: Characteristics of Arcs and Nodes
(a) 0-100 Interval (b) 0-99 Interval
Total Network Giant Component
(c) Log-Log Specification (d) Mutual Dyads
Outdegree Indegree ∆ Mean
Data Source: Hoppenstedt KSD. Figure (a) is the distribution of link weights in the [0, 100] interval and Figure (b) the corresponding [0, 99] interval. Figure (c) shows the indegree and outdegree distribution of nodes. Figure (d) reports the difference of capital weights (∆Capital Weights) for all mutual links. Each value at the abscissa is the upper threshold of a 10%-interval. For example, there are 48 links for which WAB − WBA < 10% where 10% is the upper threshold of the [0, 10) interval and WAB is the weight from vertex A to vertex B. Personally liable partners are excluded in Figure (d).
14 Other often debated shareholder structures are pyramids, also called trees and forests. 28 Due to the low absolute number of cross-holdings and due to many asymmetric dyads, tree structures should be likely to emerge. La Porta et al. (1999) as well as Faccio and Lang (2002) report the tree structures as a prevalent company structure in many developed countries. To compare the relevance of tree structures found in previous results with our data set, we can rely on the 021D and 021U triads. They represent small local trees probably embedded in large forests and also hint at the degree of centrality among nodes. The number of 021D and 021U triads is large in absolute as well as relative terms. Both triad types are observed more often than expected. 29
Hence, our statistics confirm well-known results but condenses company information in simple macro measures. Given the high number of 021D and 021U triads and the overall impression of the total German company network indicates that forests are an important structure in our data base. The emergence of these forests is often interpreted as evidence for the balance of power in company networks. Correspondingly, corporate control is exercised in the opposite direction of the arcs. Firms and subsidiaries are (partly) controlled by parent companies or other shareholders whereas the arrows tend to the controlling unit. Such forests are well documented by La Porta et al. (1999) and enhance the control of many companies by an ultimate owner.
The pyramids enable the ultimate owner to control companies he is indirectly connected to even if he is only a minority shareholder. For instance, in Network Figure 8 the shareholders of BMW are shown. Via the Dresdner Bank Allianz Corporation has direct as well as indirect influence at BMW’s general assembly. This line of argument may explain why large equity stakes of 60% and 75% are less often observed in the giant component (compare Figure 1b) than in the total network. Possibly, corporate control via forests is easier to exert in a larger network component than in smaller ones. Hence, shareholdings and forests may be substitutes.
Additional ownership structures mentioned by Windolf and Beyer (1996) are circles and (nearly) complete cliques. Also, Kogut and Walker (2001) argued that the German corporate network consists of closely knit clusters and brokers filling structural holes between these clusters. As described above, the brokers might be ultimate owners or other central companies which hold pivotal positions in the pyramids. However, the evidence for the existence of circles such as
030C triads is weak. Although the O/E-ratio of 030C triads is large, the number of observed
15 triads is low. In contrast, there is a large number of 021C triads, which confirms that circles are often found in triad formations. Yet, the expected number of 021C triads in a random network is even larger, such that the existence of circles in triads can be interpreted as a statistical artefact.
Similarly, 210 and 300 triads representing clusters and (nearly) complete cliques are infrequently observed. Hence, it is reasonable to conclude that the impact of circles, cliques, and clusters on the overall structure is moderate and at least for our data set it seems implausible to call circles or cliques a basic ownership structure. In contrast, network patterns discussed in mainstream economics journals focusing on trees and cross-holdings are prevalent.
III.C Sub-Networks
Until now, we have concentrated on information regarding the full company network. The data description is completed by turning to the analysis of subnetworks which only take into account capital linkages and related firms above certain weight thresholds. Table IV summarizes different network measures for the full network and sub-networks. Each of the three sub-networks is reduced to links with weights above 24%, 49%, or 74%. In all sub-networks the disproportionate number of 021U triads emerge again, whereas the 021D triads are less frequently observed than in a random network having the same number of vertices and arcs. These findings also suggest that the balance of power is funnelled 30 into a small number of companies which are the nodes pointed to by the arrows in the 021U triads. These companies might be the brokers mentioned in Kogut and Walker (2001) or the apex of the pyramids mentioned in La Porta et al. (2002), Claessens (2000), Attig et al. (2003), among others, which are able to coordinate different developments in their subsidiaries and, hence, occupy a strategic position which allows control of local parts of the network. Again, in all four networks 021C triads are less often observed than expected. This underpins the fact that circles are formed incidentally and cannot be seen as a power enhancing mean.
Another important feature can be read off Table IV. The number of components increases when
financial linkages below the three thresholds 24%, 49%, and 74% are ignored. The number of
16 large components above fifty nodes decreases continuously, whereas the number of components having more than five or twenty nodes first increases if we take no account of financial links below 24% but then also declines if further thresholds are considered. The giant component in the total network contains 1626 nodes and 2271 arcs. The distribution of capital weights in the giant component is similar to the distribution in the total network. Except as already mentioned, blockholdings of about 60% and 75% are found relatively seldom in the giant component, whereas in the other components these values are relatively often observed. Unsurprisingly, the giant component is quickly decomposed into smaller pieces if low weighted links are disregarded.
The giant components of all sub-networks are shown in Network Figures 2 to 4. The giant component of the sub-network containing only equity stakes above 24% consists almost completely of energy companies such as E.ON, RAG, Vattenfall 31, and others. Additionally, many public utilities are part of this sub-network. The giant component of the second sub-network containing only equity stakes above 49% is mainly a Siemens-Bosch network - one of Germany’s large technology companies - and the giant component of the 74% sub-network is an Aldi network where the Siepmann Stiftung is the center of a star. Network Figure 5 shows the complete Aldi network in which other foundations, personal liable partners, etc. are also included.
[Insert Table III about here]
[Insert Table IV about here]
17 IV Important Companies
IV.A Central Nodes in the Global Network
Here we continue the explorative analysis of the previous Section and identify the power of each corporation. The power is measured by a standard network measures the indegree closeness
32 centrality. The indegeree closeness centrality CCi of company i is defined as
|NC | |NC | CC = i i (1) i |AC| P d(i, j) j∈NCi where NCi is the set of companies which are part of the network component i belongs to, AC is the set of all companies, the bars indicate cardinality of a set, i.e. |AC| = 2784 for our data set, and d(i, j) is the distance, i.e. the length of the shortest path, between companies i and j in the same network component. 33 Companies which are closely connected to others can impact upon these companies since we take into account indegrees only. In contrast, a company which has no other equity stakes has an indegree closeness centrality of zero. Hence, it is reasonable to assume that companies with a larger CCi are more powerful than companies having a smaller centrality. 34 The closeness centrality is readily calculated and can therefore enhance the literature on company concentration (cf. Claessens et al. (2000), Faccio and Lang (2002), Attig et al. (2003),
Chapelle and Szafarz (2005), among others). 35
Unfortunately, as nearly all centrality measures also the indegree closeness centrality is not well-grounded on economic reasoning. 36 To the best of our knowledge, only Bonacich’s (1987)
Pn P+∞ k power index of company i defined as j=1 k=0 gij where j indicates all other n−1 companies,
0 < gij ≤ 1 shows that company i holds directly or indirectly g% of all shares from company j, and k is the path of length. 37 Ballester et al. (2006) have shown that the power index can be interpreted as the result of a Nash equilibrium if a quadratic utility function is supposed.
However, for our purposes the applicability of the power index is unsuitable. For instance, investors holding 75% of a company have similar control rights than investors who are the only shareholder of a company. However, the power index measure implies that a 100%-holding is
18 much more powerful where this difference in power increases with k. Even more questionable is the usefulness of the index if we compare a 100%-holding with a 10%-holding.
In the network literature, there is a whole spade of centrality measures which try to identify very important or powerful vertices. We have chosen the indegree closeness centrality because it ignores any weights and only takes into account whether there is an ownership stake or not. This choice is at least partly in accordance with definitions of the largest shareholder in the literature.
For example, Shleifer and Vishny (1997) propose to define investors holding at least 10% as large shareholders. Lech (2002) proposes a somewhat higher threshold around 25%. As shown above, companies holding stakes above these thresholds also have special rights which allow to control the management more effectively. Hence, the regulations laid down in the German shareholder act might also suggest a similar definition.
Furthermore, even a non-blockholder holding only a few percent or per mille of all shares of a company might be powerful. Since planning an investor planning an acquisition of a company might either face an opponent trying to hinder the investor activities or in the opposite case enables the investor to purchase these shares in an OTC transaction without revealing information to the market and, thereby, stock market price are not boosted. Therefore, the definition of the indegree closeness centrality might be an appropriate indicator for the power of companies. A more sophisticated centrality measure might be preferable, however, so far it is unavailable. 38
In Table V possibly important nodes are ranked by the indegree closeness centrality of nodes.
Two different rankings are shown. The full sample ranking includes all observations whereas the reduced sample ranking focuses on parent companies only and thereby, focuses on companies and discounts subsidiaries as well as state entities and individuals. The indegree statistic measures only the number of links to a company, i.e. counts the number of equity stakes a company has in other companies. As in previous studies, many insurance companies are among the most central companies. In particular, corporations such as Allianz, M¨unchner R¨uck, and Ergo as well as many subsidiaries of these companies are found. For instance, Allianz Subalpina 39 is a
19 98.003% subsidiary of RAS Riunione Adriatica di Sicurt`aS.p.A. which is a 76.34% subsidiary of the Allianz concern. Similarly, D.A.S., Hamburg-Mannheimer SV, and Victoria Versicherung are all part of the Ergo concern. For details, see the Ergo network imaged in Network Figure
11. Other frequently found industries are banks, energy suppliers, wholesale and retail firms.
Among the banks there are large German banks but there are also many foreign competitors from Italy such as UniCredito, the parent company of the Bayerische Hypo- und Vereinsbank, and Mediobanca. Details about the investment bank Mediobanca, its ownership structure, its power in Italia, and its recent role in the hostile take over of Telecom Italia is provided by
Kruse (2005) and Meoli et al. (2006). Japanese banks such as Japan Trustee Services Bank, The
Mitsubishi Trust & Banking Corporation, and Sumitomo Mitsui Banking Corporation can also be found. Japanese banks tend to cluster in local company networks called keiretsus (cf. Lincoln et al. (1996) and Lincoln and Gerlach (2004)). 40 The Japanes banks have a high number of linkages among firms but are not among the most central companies. In contrast, Italian banks and insurance companies exhibit a high closeness centrality and, therefore, might be more influential on the German economy than companies from other countries.
The national energy market is dominated by E.ON, EnBW, RWE, and Vattenfall 41. All local-operating German energy companies have to use the power grid of these four companies covering the whole German state. Each of the four big energy players covers a certain geographical area and much smaller competitors operating on a local basis have to use the power grid of one of these companies. Therefore, the big four energy companies are at least within their industry relatively powerful and except Vattenfall is not listed among the top 30 in the reduced sample ranking in Table V. Parts of their ownership structure is shown in Network
Figure 2 which stresses the strong interconnectedness among many energy companies as well as their close relationships to public utilities and cities.
Table V contains also foundations called Markus Stiftung and Luks Stiftung. Both are part of the
Aldi concern which are one of Germany’s and Europe’s largest retailers.1 The beautiful Network
1Actually, there are two concerns Aldi Nord and Aldi S¨ud.
20 Figure 5 for Aldi is an isolated network component in the total network. Additional information about the Aldi network can also be found in Network Figure 4, i.e. the giant component of the total network where links below 74% are eliminated. The owners of both companies are the brothers Karl and Theo Albrecht and are the richest Germans 42 Accordingly, the entity
‘Familie Albrecht’ is also related to this company network.
Finally, the French state - Republik Frankreich - is one of the entities exhibiting a high closeness centrality in the full sample ranking. It is well known that the French state is a large blockholder in large French companies. However, we were quite surprised to learn that this entity is found to play a central role in the German company network, too. Among the direct links, only a 50% holding of the ‘Stiftung Centre Culturel Franco-Allemand de Karlsruhe’ and a 49.02% holding of the ‘Internationale Mosel’ are reported. Both participations are rather unimportant for the overall network. Industrial relations between German and French companies are probably essential since the French state directly impinges on EADS, France T´el´ecom, Gaz de France, and Renault. Furthermore and even more importantly, there are several indirect relations to
financial corporations. The French state holds a 77.69% stake at GAN, and this company is a shareholder of the Italian Mediobanca, which is also among the most central banks. Mediobanca has, as shown in Network Figure 9, a strategic cross-holding with the Commerzbank and, as shown in Network Figure 6, is also an indirect shareholder of the AMB Generali Holding via
Assicurazioni Genarali. Finally, there is a seven-distance relationship with the AXA Konzern which contributes to the high centrality the French state exhibits in the German company network. The following seven-distance path is imaged in Network Figure 7.
AXA Konzern AG −→ AXA S.A. −→ Les Ateliers de Construction du Nord de la France S.A. −→ Eurazeo SA −→ Cr´editAgricole S.A. −→ Assurances G´en´eralesde France S.A. −→ C.D.C. C´asse des D´epˆotset Consignations −→ Republik Frankreich
[Insert Table V about here]
21 Figure 2: Distribution of Indegree Closeness Centrality
(a) Full Sample (b) Reduced Sample
Own Source. InClos is the variable name of indegree closeness centrality. The abscissa is restricted to values below 1.5. As shown in Table V, there are only two centrality values above this threshold. All indegree closeness centrality statistics are multiplied by 10−2.
IV.B Analysing the Centrality Concept
In this subsection we identify several factors which are related to the centrality of all firms. 43
The relationship of the left-hand-side variable InClos, measuring the indegree closeness centrality of firms and covariates, is based on two samples. The first sample includes variables which are observed for all companies, i.e. 2784 observations are available. The second sample - also called reduced sample - has a larger number of covariates but reduces the non-missing observations to
987. Figure 2 shows the distribution of InClos for both samples.
Hypotheses
Table VI describes 24 right-hand side variables included in the estimation below. The
Sign-column shows expected signs for each explanatory variable. For NET MoG we expect a positive sign since nodes of larger network components typically exhibit higher indegree closeness centrality. In contrast, we expect a negative sign for firms having a turnover above
1 billion euro. These companies are often only operational entities controlled by holdings and
22 other shareholders who are not involved in management decisions but have a great impact on
firm strategies. We also expect that large firms are more central than smaller ones. Hence, positive signs are allocated to LF Inc, List, and ACC Tot. We assume that all other legal forms have a negative impact upon the indegree closeness centrality since they indicate smaller firms, personally liable partners, or states entities. We also include a legal form indicator variable for missing observations to check whether important information may be contained there.
The results of Table V suggests positive coefficients for French, Italian, and Japanese companies.
Companies from the United Kingdom and the United States outnumber all other countries, but only little evidence for a high centrality of UK or US firms is found. Therefore, negative signs are assumed. The centrality measure in Figure 2 implies that most firms are unimportant for the whole network, whereas only a few are powerful. Since most vertices are German companies, a negative sign for COU Ger is expectable.
Banks and insurance companies were found to be central corporations in Germany (cf. H¨opner and Krempel (2004)). The public utility companies described above might also be powerful.
Hence, positive signs are expected for the first three industries mentioned in Table VI. Other industries may be less involved in the corporate company network. In contrast to these industries, negative signs for the manufacturing industry and trade industry are in accordance with our expectations. We also assume a positive sign for the regressor variable Multi since
firms offering various products may have stronger incentives to be interwoven with many other companies. Finally, higher profits as well as strong equity positions measured by ACC Pro and
ACC Equ should both positively affect the probability of acquiring other firms or expand a business and are likely to increase the centrality of a company. 44
[Insert Table VI about here]
Econometric Models
In Table VII results of the least squares regression are reported where we regress the indegree closeness centrality upon firm characteristics (see Table VI for variable names). Table VII
23 contains two Sub-Tables 7a and Sub-Table 7b. The first table reports coefficients and p-values of full-sample regressions, i.e. only firm characteristics being observed for all 2784 companies in the network are included. Equation 2 shows the estimation of column OLSA1.
InClos = β0 + β1NET + β2LF + β3IND + β4COU + β5Multi + β6List + u (2) where variable names in capitals indicate vectors (containing all variables of each variable group) and u is the error term. Sub-Table 7b contains also coefficients of accounting variables ACC being observed for only 987 German companies. Hence, the estimation results shown in column
OLSB1 are based on Equation 2 where the country vector COU is excluded but equity capital, balance sheet total, and annual net profit are inserted. Using the full variable set in columns
OLS·1 – the dot in the subscript here is used to indicate that the statement holds for both the full sample and the reduced sample – is appropriate due to the low degree of multicollinearity being found among indicator variables. In contrast, the correlations among accounting variables themselves are large enough to affect estimation results, as shown below. 45
In the second column of each Sub-Table we report the results of a stepwise regression which repeatedly decrements all insignificant variables until 5%-significant variables having coefficients above 0.05 in absolute value are left over. One disadvantage of our approach is that the indicator variables only measure average effects for each group. Hence, the centrality difference between banks and insurers in France is the same as between banks and insurers in Italy. Furthermore, the distribution of centrality shown in Figure 2 indicates a nonlinear relationship similar to a hyperbola. Therefore, we can assume that indegree closeness centrality increases more sharply if an already fairly central company adds a power-enhancing characteristic than if a peripheral company adds the same characteristic. A simple solution to take into account this form of nonlinearity is a semi-log specification in a linear model. However, this specification is not applicable due to company centralities of zero. Instead, nonlinear least squares is applied to
24 Equation 3 and Equation 4.
InClos = exp(β0 + β1NET 597 + β2IND Ins + β3IND Ban + (3)
β4COU F ra + β5COU Ita + β6COU Jap + β7List) + A
InClos = exp(β0 + β1NET 597 + β2IND Ins + β3IND Uti + (4)
β4ACC Equ + β5ACC T ot) + B
where all variables are scalars, exp(.) indicates the exponential function, and A and B are error terms. Regression results of Equation 3 are given in column NLSA whereas column NLSB reports results of Equation 4. In each estimation only significant variables which are left over in the linear stepwise regression are used as regressors in the nonlinear estimation.
Regression Results
The ordinary least squares regressions in columns OLS·1 and OLS·2 are discussed first. In particular, the accounting variables may be endogenous such that we would like to account for this problem. However, the available data set contains no reasonable instruments since all additional variables such as number of employees, further balance sheet information, and others may directly affect the dependent variable as well as the endogenous one. For each of the indicator variable groups the reference group are the other companies. The network variables are statistically and economically significant and have the expected signs. Firms which are a member of the giant component have a higher centrality, whereas industrial enterprises having large turnovers above one billion euro tend to have smaller centrality measures than average
firms. Four out of five legal form variables confirm our expectations. Contrary to expectations,
LF Inc has a negative sign. In regression a, the coefficient is statistically insignificant whereas it is highly significant in the reduced-sample regression. This slight difference might be caused by correlation between List and LF Inc of 0.388 in regression a and 0.517 in regression b. This line of argument is also underpinned by the observation that List is an important variable in the full-sample regression a and not included in column OLSb2 and NLSb. In regression b no coefficient is available for LF PP since accounting information excludes private individuals.
25 The variable LF Mis indicates the missing observations. In regression a no important influence is measured whereas in b the coefficients are statistically significant. However, in the first regression 131 observations are labelled as a missing variable, whereas only two are left in the reduced sample.
Among the coefficients of industry variables the largest values are observed for insurance companies. This confirms our results from the previous Sub-Section where these companies are among the most central companies. At first sight, the results for banks are mixed. In the full-sample regression significant positive coefficients are found, however, no higher centrality can be reported in the reduced-sample regression. This is substantiated by the fact that banks have higher balance sheet totals than non-banks. Excluding the variable ACC Tot and inserting the bank indicator variable results in a 1%-significant coefficient of 0.185. 46 Therefore, we confirm the result of Pappi et al. (1987) and H¨opner and Krempel (2004), i.e. that banks are still among the most powerful German companies. Expectations are also confirmed with respect to other industry variables. However, only IND Man is significant at the 5% level in column
OLSb1. All other coefficients have the assumed sign but are insignificant.
Similarly, the signs and sizes of country variable coefficients in the full-sample regression correspond to expectations for France, Italy, and Japan. The strongest impact is found for
Italy. The coefficient for the United States is negative, as assumed, but insignificant. For the United Kingdom and Germany results are not in accordance with expectations. In fact, for Germany the coefficient is also significant but the overall impact on closeness centrality is relatively small.
Two of the three accounting information have the expected sign such that a higher equity and a higher balance sheet total increases the centrality of the firm. In interpreting the coefficients take into account that ACC Equ is measured in million euros whereas AC Tot is measured in billion euros. Hence, given the coefficients of 0.007 and 0.400 in OLSB2 the equity variable seems to be more important. Company earnings seems to unimportant in determining the power of
firms.
26 Finally, we found a positive and significant relationship for the variables Multi and List. But a strong influence can only be measured for List in regression a, whereas Multi is dropped in the stepwise regressions OLS·2. Hence, not only large but also listed companies are more central.
The coefficients of nonlinear least squares estimation strengthen the results of the OLS regression.
All results with respect to sign and magnitude are confirmed. To compare the magnitudes of characteristics on centrality the coefficients must be plugged into Equation 3 and Equation 4.
Then the fitted centrality for Italian banks is exp(−2.093 + 0.399 + 1.319) = 0.687, whereas the centrality based on coefficients in OLSa2 is 0.606. French insurance companies using the results reported in column NLSa is exp(−2.093 + 0.842 + 0.547) = 0.495, which is close to closeness 47 centrality based on coefficients in OLSa2 is 0.506. Hence, our results seem quite robust to the nonlinear specification.
[Insert Table VII about here]
27 V Conclusion
Until now, researchers investigating ownership structures have been content with analysing small local company settings. The notion of patterns called pyramids or cross-holdings can be found over and over again in the existing literature. However, they represent nothing else than local network formations and are, of course, embedded in larger network structures. We believe that future analyses of ownership structures will be enhanced by network tools. For instance, network statistics might offer new variables such as centrality measures, distance to ultimate owners, etc. These variables might lead to new insights on the impact of ownership structures on
firm performance. In particular, the continuation of company network analysis is attractive for researchers given the lack of data in the past and the huge data availability today and in coming years. Detailed data sets make it possible for researchers as well as consultants to perform much more detailed firm policies and, hence, take into account the firm-specific environment and dependence structures of companies. Following the adoption of the network methods, the
German as well as the global corporate control system can be analysed in greater detail. Hence, we also believe that future researchers of company networks reviewing today’s state of research will conclude that the literature was still in its infancy, since even in this article only large companies and their shareholders are taken into account.
We show that the global description of company networks is possible by analysing ownership structures among German companies in 2006. The financial linkages of a huge unique data set containing 2784 single companies were constructed and described. Several statistics - standard in the social network literature - were applied to discover general features of the company network. From our point of view, one major highlight is the MAN-classification scheme, offering a micro-macro perspective which simplifies both specific firm analysis as well as country-specific or global analyses of ownership structures. After the description of certain structural properties a centrality measure, the indegree closeness centrality, was calculated for all vertices. Finally, the explanation of the centrality vector was performed by applying standard econometric techniques.
Our primary results show that most central German companies are still banks and insurance
28 companies. Given the results of Agarwal and Elston (2001) as well as Dittmann et al. (2005) that bank-controlled firms have not been able to outperform in the past our results might be interpreted as an undesirable network characteristic. Another interesting result is the high degree of internationalisation detected in the company network. Today, large German firms are multinational corporations themselves or are often strongly connected to other non-German multinationals. It is reasonable to assume that this fact is a major difference to earlier networks.
We found that the UK and US firms in the German company network outnumber firms of other nations, although most Anglo-Saxon firms are less central than firms from other nations.
In particular, Italian corporations, but also French and Japanese companies, occupy central positions in the German corporate system. The results of the MAN-classification scheme
(cf. Figure 3) indicate that especially cross-holdings and pyramids are the most common triad formations in the German company network. Other formations such as circles, which are found relatively often in absolute terms, are formed incidentally and are less often observed than in a random network. The importance of pyramids is also underpinned by the observation that in the giant network component which contains 1626 vertices the number of financial linkages with weights of about 60% and 75% is small, whereas in the total network there are many more such capital weights.
Finally, we turn our attention to the methodology applied and add a remark regarding the applied methodology. The start of the network literature is often traced back to Moreno (1934)
- incidentally, at the same time Berle and Means (1932) initiated the discussion on separation of ownership and control. Hence, today after seven decades of research, there are much more elaborated network concepts than the ones applied. However, we deliberately stick to well-known but also well-established network statistics due to their simplicity. The use of more modern network techniques can be applied in future research. For instance, network researchers are on the verge of understanding network regression methods applicable to highly interdependent data
(for an introduction to these new developments cf. Snijders (2005)). This research may open up new possibilities in social network analysis and, hopefully, will also contribute to the company network literature.
29 VI Appendix
[Insert Table VI about here]
[Insert Table IX about here]
[Insert Table X about here]
30 Figure 3: All Triads in the MAN-Classification Scheme
Source: “The triad isomorphism classes (with standard MAN labeling)”, Wasserman and Faust (1994), p. 566. The first three number counts the number of mutual dyads M, asymmetric dyads A, and null dyads N. The letter behind the number distinguishes otherwise identical triad formations from each other: D=Down, U=Up, C=Cycle, and T=Transitive.
31 References
Agarwal, R., and J. A. Elston, 2001, Bank-Firm Relationships, Financing, and Firm Performance
in Germany, Economic Letters 72, 225-232.
Allen, J. W., and G. M. Phillips, 2000, Corporate Equity Ownership, Strategic Alliances, and
Product Market Relationships, Journal of Finance 55, 2791-2815.
Atanasov, V., 2005, How much value can blockholders tunnel? Evidence from the Bulgarian
mass privatization auctions, Journal of Financial Economics 76, 191-234.
Attig, N., K. Fischer, and Y. Gadhoum, 2003, On the Determinants of Pyramidal Ownership:
Evidence on Dilution of Minority Interests, Working Paper.
Ballester, C., A. Calv´o-Armengol, Y. Zenou (2006), Who’s Who in Networks. Wanted: The
Keyplayer, Econometrica, 74, 5, 1403-1417.
Barab´asi,A.-L., and R. Albert, 1999, Emergence of Scaling in Random Networks, Science 286,
509-512.
Beauchamp, M. A., 1965, An improved index of centrality, Behavioral Science 10, 161-163.
Bebchuk, L. A., R. Kraakman, and G. Triantis, 2000, Stock Pyramids, Cross-Ownership and
Dual Class Equity: The Mechanisms and Agency Costs of Separating Control From
Cash-Flow Rights, in R. Morck, ed.: Concentrated Corporate Ownership (University of
Chicago Press, Chicago, IL).
Becht, M., and E. B¨ohmer,2003, Voting control in German corporations, International Review
of Law and Economics 23, 1-29.
Becht, M. and B¨ohmer, E. (2001), Ownership and voting power in Germany, in F. Barca and M.
Becht (eds.), The control of corporate Europe, Oxford: Oxford University Press, 128-153.
Becht, M., and A. R¨oell,1999, Blockholdings in Europe: An international comparison, European
Economic Review 43, 1049-1056.
Becht, M., and E. B¨ohmer (1997), The transparency of ownership and control in
Germany, in European Corporate Governance Network. The separation of ownership
and control: A survey of 7 European countries. Preliminary report to the
European Commission (Vol. 3). Brussels: European Corporate Governance Network
32 (http://www.ecgn.org/ecgn/euprelimreport.htm).
Berle, A., and G. Means, 1932, The Modern Corporation and Private Property, New York,
Macmillan.
Bertrand, M., P. Mehta, and S. Mullainathan, 2002, Ferreting out Tunneling: An Application
to Indian Business Groups, Quarterly Journal of Economics 117, 121-148.
Bonacich, P. (1987), Power and Centrality: A Family of Measures, American Journal of
Sociology, 92, 1170-1182.
Borgatti, S. P. (2003), The Key Player Problem, in Dynamic Social Network Modeling and
Analysis: Workshop Summary and Papers, ed. by R. Breiger, K. Carley, and P. Pattison.
Washington, DC.
Burkart, M., F. Panunzi, and A. Shleifer, 2003, Family Firms, Journal of Finance 58, 2167-2201.
Chapelle, A., and A. Szafarz, 2005, Controlling firms through the majority voting rule, Physica
A 355, 509-529.
Claessens, S., S. Djankov, and L. Lang, 2000, The Separation of Ownership and Control in East
Asian Corporations, Journal of Financial Economics 52, 81-112. de Nooy, W., A. Mrvar, and V. Batagelj, 2005. Exploratory Social Network Analysis with Pajek
(Cambridge University Press).
Del Guercio, D., and J. Hawkins, 1999, The motivation and impact of pension fund activism,
Journal of Financial Economics 52, 293-340.
Dittmann, I., E. Maug, and C. Schneider, 2005, Bankers and the Performance of German Firms,
Working Paper, University of Mannheim.
Edwards, J. S. S., and A. J. Weichenrieder, 2004, Ownership Concentration and Share Valuation,
German Economic Review 5, 143.
Faccio, M. and L.H.P. Lang, 2002, The ultimate ownership of Western European corporations,
Journal of Financial Economics 65, 365-395.
Fee, C. E., C. J. Hadlock, and T. Shawn, 2006, Corporate Equity Ownership and the Governance
of Product Market Relationships, Journal of Finance 61(3), 1217-1251.
Fohlin, C., 2005, The History of Corporate Ownership and Control in Germany, in R. K. Morck,
ed.: A History of Corporate Governance Around the World: Family Business Groups to
Professional Managers (University of Chicago Press, Chicago, IL).
33 Franks, J. R., and C. Mayer, 2001, Ownership and Control of German Corporations, Review of
Financial Studies 14, 943-977.
Franks, J. R., C. Mayer, and H. F. Wagner, 2005, The Origins of the German Corporation -
Finance, Ownership and Control, ECGI - Finance Working Paper No. 110/2005.
Freeman, L.C., 1979, Centrality in affiliation networks, Social Networks, 19, 157-191.
Gill, P. S., and T. B. Swartz, 2004, Bayesian analysis of directed graphs with applications to
social networks, Applied Statistics 53, 249-260.
Goergen, M., M.C. Manjon Antolin, and L. Renneboog, 2004, Recent Developments in German
Corporate Governance, ECGI, Finance Working Paper No. 41.
Godsil, C., and G. Royle, 2001. Algebraic Graph Theory, Graduate Texts in Mathematics
(Springer Verlag).
Gorton, G., and F. A. Schmid, 2000, Universal banking and the performance of German firms,
Journal of Financial Economics 58, 29-80.
Heinze, T., 2004, Dynamics in the German System of Corporate Governance? Empirical
Findings regarding Interlocking Directorates, Economy & Society 33, 218-238
Holland, P. W., and S. Leinhardt, 1970, A method for detecting structure sociometric data,
American Journal of Sociology, 76, 492-513.
H¨opner, M., and L. Krempel, 2004, The Politics of the German Company Network, Competition
and Change 8, 339-356.
Johnson, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer, 2000, Tunneling, American
Economic Review 90, 22-27.
Kammerath, J. (1999), Methodische und empirische Grundlagen der quantitativen Erfassung
kontrollierender Verbindungen deutscher Unternehmen mit ihren direkten und indirekten
Eignern, Report commissioned by the Ifo-Institute, Munich.
Kogut, B., and G. Walker, 2001, The Small World of Germany and the Durability of National
Networks, American Sociological Review 66, 317-335.
K¨oke, F. J., 1999, New evidence on ownership structures in Germany, ZEW Working Paper No.
99-60, Mannheim.
Kosch¨utzki,D. K., A. Lehmann, L. Peeters, S. Richter, D. Tenfelde-Podehl and O. Zlotowski,
2005, Centrality Indices, in Ulrik Brandes and Thomas Erlebach ed.: Network Analysis,
34 Lecture Notes in Computer Science (Springer Verlag).
Kruse, T. A., 2005, Ownership, Control and Shareholder Value in Italy: Olivetti’s Hostile
Takeover of Telecom Italia, Finance Working Paper 83, University of Arkansas.
La Porta, R., F. Lopes-de-Silanes, A. Shleifer, and R. W. Vishny, 1998, Law and finance, Journal
of Political Economy 106, 1113-1155.
La Porta, R., F. Lopes-de-Silanes, and A. Shleifer, 1999, Corporate Ownership Around the
World, Journal of Finance 54, 471-517.
La Porta, R., F. Lopes-de-Silanes, A. Shleifer, and R. W. Vishny, 2002, Investor Protection and
Corporate Valuation, Journal of Finance, 57, 1147-1170.
Leech, Dennis (2002): An Empirical Comparison of the Performance of Classical Power Indices.
In: Political Studies, Vol. 50, pp. 1-22.
Lemmon, M. L., and K. V. Lins, 2003, Ownership Structure, Corporate Governance, and Firm
Value: Evidence from East Asian Financial Crisis, Journal of Finance 58, 1145-1168.
Lincoln, J. R., M. Gerlach, and C. Ahmadjian, 1996, Keiretsu networks and corporate
performance in Japan, American Sociological Review 61, 67-88.
Lincoln, J. R. and M. Gerlach, 2004. Japan’s Network Economy: Structure, Persistence, and
Change (Cambridge University Press).
Maury, B., 2005, Family ownership and firm performance: Empirical evidence from Western
European corporations, Journal of Corporate Finance 12, 321-341.
Meoli, M., S. Paleari, and G. Urga, 2006, When controlling shareholders “live like kings”. The
case of Telecom Italia, Working Paper.
Mitzenmacher, M., 2003, A Brief History of Generative Models for Power Law and Lognormal
Distributions, Internet Mathematics 1, 226-251.
Miyajima, H., and F. Kuroki, 2005, The Unwinding of Cross-Shareholding in Japan: Causes,
Effects, and Implications, RIETI Discussion Paper Series 05-E-006.
Moreno, J. L., 1934, Who Shall Survive? Foundations of Sociometry, Group Psychotherapy and
Sociodrama. Nervous and Mental Disease Publishing Company, Washington, DC.
Newman, M. E. J., 2001, Scientific collaboration networks. II. Shortest paths, weighted networks,
and centrality, Physical Review E 64.
Nowak, E., O. Ehrhardt, and F. M. Weber, 2006, ’Running in the Family’: The Evolution of
35 Ownership, Control, and Performance in German Family-Owned Firms, 1903-2003, Swiss
Finance Institute Research Paper No. 06-13.
Pappi, F. U., P. Kappelhoff, and C. Melbeck, 1987, Die Struktur der
Unternehmensverflechtungen in der Bundesrepublik. Eine Blockmodellanalyse der
Personal- und Kapitalverflechtungen zwischen den gr¨oßten Unternehmen, K¨olner
Zeitschrift f¨ur Soziologie und Sozialpsychologie 39, 693-717.
Prigge, S. (2007), The Performance of Measures of Shareholder Influence, mimeo.
Sabidussi, G., 1966, The centrality index of a graph, Psychometrika 31, 581-603, 1966.
Shleifer, A. and R. Vishny (1997), A survey of corporate governance, Journal of Finance, 52,
737-784.
Snijders, T. A. B., 2005, Models for Longitudinal Network Data, in Peter Carrington, John
Scott, and Stanley Wasserman, ed.: Models and Methods in Social Network Analysis
(Cambridge University Press).
Terberger, E., 2003, The German financial system: great institutional change and little effect?,
Journal of Theoretical and Institutional Economics 159, 707-716.
Watts, D., and S. Strogatz, 1998, Collective Dynamics of ’Small World’ Dynamics, Nature 393,
440-442.
Villalonga, B., and R. Amit, 2006, How do family ownership, control and management affect
firm value?, Journal of Financial Economics 80, 385-417.
Wasserman, S., and K. Faust, 1994. Social Network Analysis: Methods and Applications
(Cambridge University Press).
Windolf, P., and J. Beyer, 1996, Cooperative Capitalism, British Journal of Sociology 47,
205-231.
Windolf, P., 2002. Corporate Networks in Europe and the United States (Oxford University
Press).
Windolf, P., 2005, The Emergence of Corporate Networks in Germany and the United States
1896-1938, Working Paper.
Yafeh, Y., and O. Yosha, 2003, Large Shareholders and Banks: Who Monitors and How?,
Economic Journal 113, 128-146.
36 Endnotes
1. Cf. also Goergen et al. (2004) who provide a review of German corporate governance system.
2. Agency costs might harm whereas increased monitoring efforts can create efficiency, evidence for the latter argument evidence is found by Yafeh and Yosha (2003), Gorton and Schmid (2000) and mixed results are found by Del Guercio and Hawkins (1999)).
3. Edwards and Weichenrieder (2004) provides us with an econometric method to distinguish detrimental and beneficial effects of large shareholders.
4. Authors’ note.
5. “Our principal contribution is to find wherever possible the identities of the ultimate owners of capital and of voting rights in firms, so when shares in a firm are owned by another company, we examine the ownership of that company, and so on. For most countries, this is the only way to understand the relationship between ownership and control. These data enable us to address, in a comparative perspective, four broad questions... .” La Porta et al. (1999), p. 472.
6. For instance, the equity stakes of the largest one hundred companies are investigated in biennial reports of the German Monopolkommission. Publications of H¨opner and Krempel
(2004) are often based on this data set.
7. For simplicity, all non-financial companies are called industrial companies.
8. The Economist, Dec 5th 2002, print edition.
37 9. Besides a higher degree of internationalisation, concentration within an industry might also be an alternative explanation for the dissolution process among large companies. Brisk competition might force cooperation among firms. Hence, links are broken across an industry whereas the interlocking within main markets of companies is intensified. Allen and Phillips (2000) found a positive impact on operating performance in research intensive industries if blockholdings are combined with product market relationship between purchasing and target firm. Also Fee et al. (2006) investigates the impact of financial linkages among trading partners. They find that equity stakes between customers and suppliers increase the time span of trade relationships.
Given these findings, it will be interesting to investigate whether the ownership structure is intensive within industries. However, the answers to these questions are left to future research.
10. An overview of large company networks in six different countries is given in Windolf (2002).
11. The editor of one of Germany’s large business newspapers ‘Frankfurter Allgemeine Zeitung’.
12. The rewiring procedure picks company u that severs an existing link to company v and forms a new one to company w (see Watts and Strogatz (1999) for details).
13. The German name is ‘Monopolkommission’.
14. The data bank is available at www.hoppenstedt-konzernstrukturen.de
15. See Table 2 in Goergen et al. (2004) for further references.
16. Furthermore, the sometimes mentioned data source “Wer geh¨ortzu wem?” (which means
“who owns whom?”) of the Commerzbank is based on the KSD.
38 17. Financial companies are not part of this core sample since by definition financial companies have no balance sheet item called turnover.
18. Throughout the paper we use reasonable abbreviations for company names. In particular, legal forms of companies are never mentioned in the text. The legends of network figures shown in the Appendix contain full company names. For instance, BMW is called ‘Bayerische Motoren
Werke AG’ in Network Figure 8.
19. A member of the control board in corporation A cannot be member of the executive board of corporation B if an executive member of corporation B is a member of the control board of corporation A, §100(2)Nr.3 AktG (Prohibition of cross interlocks).
20. Niedersachsen is one out of 16 German states.
21. Also, the German banking industry has specific regulations. Almost all cities and communities are owners of small saving banks - called Sparkassen - which all together are larger with respect to standard bank characteristics than most listed German competitors.
22. In fact, the 192 components are weak components which take into account all companies being connected to each other independent of the direction of the arrows (strong components distinguish the direction of the arrows). See de Nooy et al. (2005) for details.
23. In all Network Figures links are classified into three groups where thicker lines stand for higher equity stakes among the firms. The thinnest lines represent equity stakes up to 10%, medium lines represent stakes from 10% up to 50%, and the thickest lines represent equity stakes from 50% to 100%.
39 24. The German company act is called Aktiengesetz and is commonly abbreviated by AktG.
25. Mathematically, the linearity is reproducible by power law or lognormal distributions.
Barab´asiand Albert (1999) show that many network data sets exhibit power laws. For a general discussion of the characteristics of these distributions and how human behaviour can produce such distributions read Mitzenmacher (2003).
26. Cross-holdings are defined as direct cross-holdings whereas K¨oke (1999) uses a broader definition which also takes into account circles of large distances.
27. In network terminology, triads are networks among three nodes and dyads are networks among two nodes.
28. In the corporate governance literature these structures are called pyramids, whereas the graph theoretical notion is tree or forest. Hence, we also use to the last notions. Cf. Godsil and
Royle (2001).
29. An overall test of independence has a χ2-value of 6 108 and, accordingly, clearly refutes the notion that the network is formed by accident.
30. This notion is introduced into the network literature by Newman (2001). It implies that all geodesic paths from one vertex to all others in a network component typically go through a very small number of adjacent vertices.
31. Vattenfall is a Swedish company.
32. Notice, that this formula deviates from the standard closeness centrality since our network consists of several components. The standard centrality definition is extended by the the first
40 fraction which controls for the number of nodes in each network component.
33. See Kosch¨utzkiet al. (2005) for definitions and more advanced centrality statistics.
34. The article by Freeman (1979) is a standard reference, although he was not the first to propose centrality concepts. Compare, for instance, Beauchamp (1965) and Sabidussi (1966).
35. Interestingly, without mentioning the term ‘network’, Chapelle and Szafarz (2005) use network techniques by applying matrix algebra to calculate ultimate owners. Note that, mathematically the notions ‘network’ and ‘matrix’ are synonyms.
36. Borgatti (2003) discusses problems in applied work which arises due to the lack of a theoretical foundation.
37. In the original work the power index also includes a scaling factor. For brevity, we omit it.
38. Furthermore, game theoretical power measures such as the Shapley-Shubik or the Banzhaf index exhibit undesirable features. Compare Prigge (2007).
39. This company holds rank 24 in the indegree closeness centrality column. The registered name is ‘Allianz Subalpina Societ`adi assicurazioni e riassicurazioni’, based in Turin.
40. Miyajima and Kuroki (2005) show that Japanese firms can be separated into two groups after the banking crisis in the nineties. The less efficient companies are still strongly connected with banks, whereas the more prosperous corporations exhibit a higher tendency to break these links.
41 41. Vattenfall is a Swedish company.
42. Their wealth is estimated at approx. 18.5 and 15.5 billion USD. See Forbes Special Report
‘The World’s Billionaires’ 03.10.2005.
43. Heinze (2004) applied the same methodology we adopted here to explain the centrality of interlocking directorates. However, we have some doubt about the validity of this method. The independence assumption prerequisite for the application of standard econometric methods is violated in the case of network data (see Gill and Swartz (2004)). Fortunately, if our doubts are unfounded, then results are viable and if our doubts are justified, then many results published in well-known journals may be error-prone since to the best of our knowledge the interdependencies among companies are always ignored. The issue of interdependence among observations is especially important since most studies focus on large companies which are often closely related in one form or another form.
44. Notice, for all variable groups the reference group always contains all other companies.
45. The correlations mentioned are ρ(ACC Tot,ACC Pro)=0.386, ρ(ACC Tot,ACC Equ)=0.411,
ρ(ACC Equ,ACC Pro)=0.767.
46. The corresponding p-value is 0.008.
47. The last value rests upon the following equation 0.114+0.248+0.144=0.506.
42 Captions of Figures Figure 1: Characteristics of Arcs and Nodes
Figure 2: Distribution of Indegree Closeness Centrality
Figure 3: All Triads in the MAN-Classification Scheme
43 Tables
44 Table I: Country Ranking Country %-share Country %-share Country %-share Germany 70.47 Cayman Islands 0.32 Bahrain 0.04 US 4.71 Norway 0.32 Cyprus 0.04 United Kingdom 3.23 Bermuda 0.25 Czech Republic 0.04 Italy 3.20 Canada 0.18 Ireland 0.04 France 3.09 UA Emirates 0.14 Iran 0.04 The Netherlands 2.91 Russia 0.14 Libya 0.04 Japan 2.73 Denmark 0.11 Monaco 0.04 Swiss 1.98 South Africa 0.11 Mexico 0.04 Luxembourg 1.36 Finland 0.07 Portugal 0.04 Austria 1.19 Hong Kong 0.07 T&C Islands 0.04 Belgium 0.93 Kuwait 0.07 Virgin Islands 0.04 Sweden 0.93 Korea 0.04 ———————– ———— Spain 0.54 Saudi Arabia 0.04 Total 100.0 Australia 0.40 Netherlands Antilles 0.04 ———————– ———— Data Source: Hoppenstedt Konzernstruktur Datenbank (KSD). The total number of companies is 2784. UAE abbreviates United Arab Emirates. Official country name of Ireland is ‘The Republic of Ireland’, and T&C Islands full name is ‘Turks and Caicos Islands’.
45 Table II: Legal Forms of Companies in our Sample Legal Form Legal Form Group in Germany #Obs %-share Private Limited Company Ltd.-Group GmbH 1023 36.75 Public Limited Company Inc.-Group AG 690 24.78 Partnership Partner-Group KG/OHG 303 10.88 Others Other-Group 69 2.48 Foundations Stiftung 35 1.26 Cooperatives e.G. 26 0.93 Civil Law Association GbR 4 0.14 Association e.V. 4 0.14 Private Individuals 368 13.22 State Enterprises 102 3.66 Missing Observations 229 8.23 Sum 2784 100.00 Data Source: Hoppenstedt KSD - data bank access is provided via www.konzernstrukturen.de. Abbreviations are listed in Table II. #Obs signify the number of observations.
46 Table III: Observed and Expected Number of Isomorphic Triads MAN-Type Observed Expected O/E ratio 003 3,582,363,243 3,582,129,437.72 1.0000653 012 9,821,853 10,288,226.42 0.95 102 227,396 2,462.40 92.35 021D 6,307 2,462.40 2.56 021U 5,872 2,462.40 2.38 021C 4,179 4,924.80 0.85 030T 473 2.36 200.42 111U 385 2.36 163.14 111D 157 2.36 66.53 201 75 0.00 dbz 030C 12 0.79 15.19 120D 11 0.00 dbz 120U 9 0.00 dbz 120C 8 0.00 dbz 210 4 0.00 dbz 300 0 0.00 dbz Sum 3,592,429,984 3,592,429,984 Data Source: Hoppenstedt KSD. MAN-Types are defined by Holland and Leinhardt (1970). M counts the mutual dyads, A the asymmetric dyads, and N the null dyads in a triad. In addition, D down, U up, T transitive, and C cycle indicate the direction of links in asymmetric dyads. Confer also Figure 3 and Table VI. The ‘O/E ratio’ is the ratio of observed number of triad types in our data set relative to the expected number of triad types in a random network model. dbz abbreviates ‘division by zero’.
47 Table IV: Importance of Capital Weights for Company Sub-Networks Threshold 0% 24% 49% 74% Companies 2784 2061 1867 1585 Arcs 3711 1771 1445 1152 021D-Triads 6307+ 223− 43− 1− 021U-Triads 5872+ 2207+ 1898+ 1009+ 021C-Triads 4179− 1054− 696− 493− Components 192 373 433 437 Component(Companies>5) 66 107 97 83 Component(Companies>20) 8 20 10 3 Component(Companies>50) 3 2 0 0 Companies in Giant Component 1626 117 40 28 Arcs in Giant Component 2271 122 40 27 Data Source: Hoppenstedt KSD. The full network has a threshold of 0%. A sub-network includes all links with weights above the threshold level. 021D-triads counts the number of triads with zero mutual, two asymmetric, one null dyad, and D indicates that both arrows point to one link, i.e. there is one shareholder with two different equity stakes (U=up, C=cycle). +(−) indicates whether the observed number of triads is above (below) the expected number of triads. Component(Companies>K) counts the number of network components containing more than K companies, where the number of network components is the number of totally disconnected network parts.
48 Table V: Most Central Companies and Entities
10−2 10−2 Full Sample Ranking Reduced Sample Ranking #Obs: 2784 #Obs: 275
Rank Company/Entity Indegree Closeness Company/Entity Indegree Closeness 1 Allianz AG 3.28 Allianz AG 3.28 2 M¨unchener R¨uck AG 3.11 M¨unchener R¨uck AG 3.11 3 Familie Albrecht 1.66 Bayer. Hypo- und Vereinsbank AG 1.04 4 Republik Frankreich 1.62 COMMERZBANK AG .95 5 UniCredito Italiano SpA 1.44 E.ON AG .88 6 Fondazione Cassa di Risparmio di Torino 1.35 Coca-Cola Erfrischungsgetr¨anke AG .74 7 AVIVA Plc 1.35 KARSTADT QUELLE AG .65 8 Fondazione Cassa di Risparmio Verona1) 1.30 RWE AG .60 9 Markus Stiftung 1.26 DEUTSCHE BANK AG .59 10 Barclays PLC 1.24 Norddeutsche Landesbank .49 11 The Capital Group Companies Inc. 1.23 Siemens AG .46 12 Capital Research & Management2) 1.23 EnBW AG .43 13 Lukas Stiftung 1.22 ThyssenKrupp AG .39 14 Assicurazioni Generali SpA 1.21 Landesbank Baden-W¨urttemberg .34 15 Ergo Versicherungsgruppe AG 1.16 Franz Haniel & Cie. GmbH .33 16 RAS Riunione Adriatica di Sicurt`aSpA 1.14 SHB AG11) .32 17 Mediobanca Banca dCF SpA3) 1.11 EDEKA ZENTRALE AG & Co. KG .32 18 Victoria Versicherung AG 1.11 Deutsche Lufthansa AG .29 19 Hamburg-Mannheimer SV AG 1.11 Deutsche Bahn AG .28 20 Fondazione Cassamarca5) 1.11 Robert Bosch GmbH .27 21 RB Vita SpA 1.11 DaimlerChrysler AG .27 22 Europ¨aische Reiseversicherung AG 1.11 REWE-ZENTRALFINANZ e.G. .27 23 Carimonte Holding SpA 1.11 RAG AG .22 24 Allianz Subalpina8) 1.11 VOLKSWAGEN AG .22 25 D.A.S. AG9) 1.11 MAN AG .21 26 Fidelity Investments Ltd 1.10 Deutsche Telekom AG .21 27 Legal & General Group PLC 1.10 Deutsche BP AG .20 28 D.A.S. AG10) 1.10 Salzgitter AG .19 29 KarstadtQuelle Lebensversicherung AG 1.10 ExxonMobil GmbH12) .19 30 DKV Deutsche Krankenversicherung AG 1.10 Bayerische Landesbank .19 Own Source: Full company names are provided to simplify identification of companies. #Obs signifies the number of observations. 1) − 12) full names are provided in Table IX in the Appendix. Note 9) and 10) have identical abbreviations but different full names. Further abbreviation: SV=Sachversicherung (property insurance). Translations: Lebensversicherung=life insurance, Krankenversicherung=health insurance, Reiseversicherung=travel insurance, Familie=family, Stiftung=foundation, Europ¨aische=European, Republik Frankreich=France. The international company name of M¨unchner R¨uck is Munich Re Group. In the reduced sample ranking, several firms of the EDEKA association have a closeness centrality of approximately 0.25 and are among the top 30. However, we exclude them since EDEKA Zentral is already considered in the ranking.
49 Table VI: Covariates available for the Explanation of Indegree Closeness Centrality
Category (abbr.) Variable Description Sign Network NET MoG Indicates firms being a member of the giant + component (NET) NET 597 Indicates firms having a turnover above 1 bill. Euro – Legal Form LF Inc Indicates incorporated companies + (LF) LF Ltd Indicates limited companies – LF Par Indicates partnerships – LF PP Indicates personally liable partners – LF Sta Indicates state enterprises/state entities – LF Mis Indicator variables for missing observations + – Industry IND Ins Indicates insurance companies + (IND) IND Ban Indicates banks + IND Uti Indicates public utility companies + IND Man Indicates manufacturing companies – IND Tra Indicates wholesale and retail companies + – Country COU Ger Indicates German firms or entities – (COU) COU Fra Indicates French firms or entities + COU Ita Indicates Italian firms or entities + COU Jap Indicates Japanese firms or entities + COU UK Indicates British firms or entities – COU USA Indicates U.S. firms or entities – Conglomerate Multi Indicates firms being active in a main industry and at + least five sub-industries Listed List Indicates firms having positive market capitalization + Accounting ACC Tot Balance sheet total + (ACC) ACC Pro Annual net profit + ACC Equ Equity Capital + Own Source: All variables of the first six categories (from the Network-category up to the Listed-category) are indicator variables and are observed for the whole sample - 2784 companies. The variables of the Accounting category is observed for 987 companies.
50 Table VII: (Non-)Linear Least Squares Estimation
Table VII contains several regression results. In each estimation the dependent Variable is InClos, the indegree closeness centrality. The Table is split into two Sub-Tables 7A and 7B. The regression results shown in the first Sub-Table are based on the full sample, whereas the regression results in the second Sub-Table is based only on 987 observations. However, the reduced sample includes also accounting information of German companies. The first two columns of each Sub-Table are estimated by ordinary least squares, whereas the last column reports results of nonlinear least squares estimation.
Sub-Table 7 a Sub-Table 7 b Full Sample Reduced Sample
Variable OLSa1 OLSa2 NLSa Variable OLSb1 OLSb2 NLSb Network Variables Network Variables NET MoG 0.043** NET MoG -0.017 (0.000) (0.136) NET 597 -0.100** -0.084** -1.120** NET 597 -0.048* -0.058** -0.527** (0.000) (0.000) (0.000) (0.015) (0.003) (0.001) Legal Form Variables Legal Form Variables LF Inc -0.046 LF Inc -0.094* (0.127) (0.012) LF Ltd -0.076** LF Ltd -0.111** (0.009) (0.002) LF Par -0.068* LF Par -0.094* (0.021) (0.011) LF PP -0.078** LF PP (0.009) LF Sta -0.061+ LF Sta -0.162* (0.068) (0.028) LF Mis -0.006 LF Mis -0.036 (0.871) (0.661) Industry Variables Industry Variables IND Ins 0.229** 0.248** 0.842** IND Ins 2.830** 2.929** 2.947** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) IND Ban 0.054* 0.066** 0.399** IND Ban 0.107 (0.029) (0.007) (0.000) (0.123) IND Uti 0.007 IND Uti 0.120* 0.139** 1.082** (0.524) (0.015) (0.009) (0.000)
51 Sub-Table 7 a continued Sub-Table 7 b continued IND Man -0.005 IND Man -0.038* (0.526) (0.012) IND Tra 0.005 IND Tra -0.008 (0.599) (0.656) Country Variables Accounting Variables COU Ger 0.029** ACC Pro -0.030 (0.005) (0.329) COU Fra 0.134** 0.144** 0.547** ACC Equ 0.010** 0.007** 0.011** (0.000) (0.000) (0.000) (0.007) (0.001) (0.000) COU Ita 0.417** 0.426** 1.319** ACC Tot 0.256** 0.400* 1.106** (0.000) (0.000) (0.000) (0.007) (0.012) (0.000) COU Jap 0.096** 0.096** 0.433** (0.000) (0.000) (0.000) COU UK 0.044 (0.111) COU USA -0.019 (0.279) Other Variables Other Variables Multi 0.051+ Multi 0.056** (0.081) (0.001) List 0.048+ 0.069** 0.443** List 0.014 (0.050) (0.005) (0.000) (0.538) Constant 0.134** 0.114** -2.093** Constant 0.182** 0.101** -2.109** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.279 0.257 0.4661) R2 0.870 0.849 0.8391) #Obs 2784 2784 2784 #Obs 275 275 275
Data Source: Hoppenstedt KSD and Hoppenstedt Annual Data Information (www.bilanzen.de). p-values in parenthesis. 1%, 5%, 10% significance levels are labelled by **, *, +. All except the accounting variables ACC Pro, ACC Tot, and ACC Equ are indicator variables. ACC Pro and ACC Equ is measured in million Euros whereas ACC Tot is measured in billion Euros. 1) R2 is not the standard goodness-of-fit since the nonlinear least squares regression contains no intercept.
52 Table VIII: Probabilities in a Random Network
Conditional Probabilities for each triad type P (003|0L) = 1 P (012|1L) = 1 P (102|2L) = 0.2 P (021D|2L) = 0.2 P (021U|2L) = 0.2 P (021C|2L) = 0.4 P (111D|3L) = 0.3 P (111U|3L) = 0.3 P (030T |3L) = 0.3 P (030C|3L) = 0.1 P (201|4L) = 0.2 P (120D|4L) = 0.2 P (120U|4L) = 0.2 P (120C|4L) = 0.4 P (210|5L) = 1 P (300|6L) = 1 Probabilities that a certain number of links is formed in a triad. 6 0 6 P (0L) = 0 p (1 − p) 6 1 5 P (1L) = 1 p (1 − p) 6 2 4 P (2L) = 2 p (1 − p) 6 3 3 P (3L) = 3 p (1 − p) 6 4 2 P (4L) = 4 p (1 − p) 6 5 1 P (5L) = 5 p (1 − p) 6 6 0 P (6L) = 6 p (1 − p) Own Source: A random network is defined as a network where each link has the same formation probability p.
Existing Links 3711 In our case p = Maximal Number of Links = 2784 2783 = 0.00047897. Therefore, the expected number of triads which include many arcs in Table III, for instance number of 300-triad, 210-triads, etc., is close to zero. 0L=zero links are formed, 1L=one link is formed,. . . ,6L=six links are formed. It holds that P (2L) = P (102) + P (021D) + P (021U) + P (021C) and similar for P (3L) and P (4L).
53 Table IX: Full Company Names abbreviated in Table V Footnote Company 1) Fondazione Cassa di Risparmio Verona, Vicenza, Belluno e Ancona 2) Capital Research & Management Company 3) Mediobanca Banca di Credito Finanziario S.p.A. 4) The Mitsubishi Trust & Banking Corporation (Mitsubishi Shintaku Ginko) 5) Fondazione Cassamarca - Cassa di Risparmio della Marca Trivigniana 6) Fidelity Management & Research Company 7) The Dai-Ichi Kangyo Bank, Ltd. (Dai-Ichi Kangyo Ginko) 8) Allianz Subalpina Societ`adi assicurazioni e riassicurazioni 9) Deutscher Automobil Schutz Allgemeine Rechtsschutz-Versicherungs-AG 10) D.A.S. Deutscher Automobil Schutz Versicherungs-AG 11) SHB Stuttgarter Finanz- und Beteiligungs AG 12) ExxonMobil Central Europe Holding GmbH Own Source.
54 Table X: Legal Forms of Companies in our Sample
Abbreviation Countries Local Name Group #Obs A/S Denmark Aktieslskab Inc. 9 AB Sweden Aktiebolag Inc. 15 AG Germany Aktiengesellschaft Inc. 374 AG & Co KG Germany Inc. 16 ASA Norway Allmennaksjeselskap Inc. 2 BV The Netherlands Besloten Vennootshap met Ltd. 44 beperkte aansprakelijkheid CV The Netherlands Commanditaire Vennootschap Partner 2 e.G. Germany eingetragene Genossenschaft Other 26 e.V. Germany eingetragener Verein Other 4 Foundation Anglo-Saxon Other 1 GbR Germany Gesellschaft des b¨urgerlichen Rechts Other 4 GmbH Germany Gesellschaft mit beschr¨ankter Haftung Ltd. 816 GmbH & Co. KG Germany Partner 213 GmbH & Co. oHG Germany Partner 16 KG Germany Kommanditgesellschaft Partner 43 KGaA Germany Kommanditgesellschaft auf Aktien Inc. 7 LLC USA Limited Liability Company Partner 22 LLP USA, UK Limited liability partnership Partner 3 LP USA Limited Partnership Ltd. 16 Ltd. UK Limited Ltd. 133 NV Belgium Naamloze Vennootschap Inc. 37 The Netherlands Naamloze Vennootschap Inc. oHG Germany offene Handelsgesellschaft Partner 4 PLC UK Public company limited by shares Inc. 31 SA Belgium Soci´et´eAnonyme Inc. 120 Brazil Sociedade Anˆonima Inc. France Soci´et´eAnonyme Inc. Luxembourg Soci´et´eAnonyme Inc. Portugal Sociedade Anˆonima Inc. Spain Sociedad An´onima Inc. SARL France Societe a responsabilite limitee Ltd. 14 Luxembourg Societe a responsabilite limitee Ltd. SAS France Soci´et´epar Actions Simplifi´ee Inc. 7 SCA France Soci´et´een commandite par actions Inc. 4 SPA Italy Societa per azioni Inc. 68 Stiftung Germany Stiftung Other 34 Own Source: #Obs signifies the number of observations in the data set.
55 Network Figure 1: Total Network Network Figure 2: 51 Giant 24 49 70 3
115 69
4 12
5
92 41 1 112 116 2 68 110 71 63 111 72 76 90 75 53 93 84 11 60 61 95 113 10 58 94 86 47 83 88 57 59 19 6 80 40 62 8 27
54 52 32 48 91 16 67 29 13 23 108 28 21 31 87 35 24 96 26 114 18 17 9 43 14 105 85 30 104 77 25 15 42 20 66 102 55 109 36 22 97 33 100 50 64 99 103 74 65 73 44 82 56 7 98 106 89 34 101 107 45 79 117
39
37 78 81
38 46 Giant 24 This is the Giant Component of the Complete Network where all links below 24% are deleted.
1 Aktien-Gesellschaft der Dillinger Hüttenwerke 42 Gasag Berliner Gaswerke AG 83 Stadtwerke Augsburg Energie GmbH 2 ARBED S.A. 43 Gasanstalt Kaiserslautern AG 84 Stadtwerke Augsburg Holding GmbH 3 Arcelor Eisenhüttenstadt GmbH 44 Gaz de France Berliner Investissements SAS 85 Stadtwerke Chemnitz AG 4 Arcelor Germany Holding GmbH 45 Gaz de France Deutschland GmbH 86 Stadtwerke Frankfurt am Main Holding GmbH 5 Arcelor S.A. 46 Gaz de France Produktion Expl. Deutschland GmbH 87 Stadtwerke Hannover AG 6 Bayerngas GmbH 47 HEAG AG 88 Stadtwerke München GmbH 7 BKB AG 48 HEAG Südhessische Energie AG (HSE) 89 Stadtwerke Regensburg GmbH 8 citiworks AG 49 Jean Lang 90 Stadtwerke Strom-/Wärmeversorgungsgesell. mbH 9 CONTIGAS Deutsche Energie-AG 50 Kreise 91 Stadtwerke Zweibrücken GmbH 10 Degussa AG 51 Landeselektrizitätsverband Oldenburg 92 Stahlwerke Bremen GmbH 11 Deutsche Steinkohle AG 52 Landeshauptstadt Hannover 93 STEAG AG 12 DHS - Dillinger Hütte Saarstahl AG 53 Landeshauptstadt München 94 STEAG Saar Energie AG 13 E.ON AG 54 Mainova AG 95 SWM Versorgungs GmbH 14 E.ON Avacon AG 55 N-ERGIE AG 96 Thüga AG 15 E.ON Bayern AG 56 Öffentliche Gebietskörperschaften 97 Thüringer Energie-Beteiligungsgesellschaft mbH 16 E.ON edis AG 57 RAG AG 98 Vattenfall (Deutschland) GmbH 17 E.ON Energie 26. Beteiligungs-GmbH 58 RAG Beteiligungs-GmbH 99 Vattenfall AB 18 E.ON Energie AG 59 RAG Coal International AG 100 Vattenfall Europe AG 19 E.ON Finanzanlagen GmbH 60 RAG Projektgesellschaft mbh 101 Vattenfall Europe Berlin AG & Co. KG 20 E.ON Hanse AG 61 RAG Saarberg GmbH 102 Vattenfall Europe Berlin Verwaltungs-AG 21 E.ON Kernkraft GmbH 62 RAG Trading GmbH 103 Vattenfall Europe Generation AG & Co. KG 22 E.ON Kraftwerke GmbH 63 RAG Verkauf GmbH 104 Vattenfall Europe Generation Verwaltungs-AG 23 E.ON Mitte AG 64 Regensburger Badebetriebe GmbH 105 Vattenfall Europe Hamburg AG 24 E.ON Netz GmbH 65 Regensburger Energie- und Wasserversorgung AG 106 Vattenfall Europe Sales GmbH 25 E.ON Nordic AB 66 Rewag Regensburger Ener.- und Wass. AG & Co KG 107 Vattenfall Europe Transmission GmbH 26 E.ON Nordic Holding GmbH 67 Rütgers GmbH 108 Versorgungs- und Verkehrsgesellschaft Hannover mbH 27 E.ON RAG-Beteiligungsgesellschaft mbH 68 Saar Ferngas AG 109 Versorgungs- und Verkehrshold. GmbH Chemnitz (VVHC) 28 E.ON Ruhrgas AG 69 Saarstahl AG 110 Verwaltungsgesellschaft RAG-Beteiligung mbH 29 E.ON Ruhrgas Holding GmbH 70 SHS - Struktur-Holding-Stahl GmbH & Co. KGaA 111 VNG - Verbundnetz Gas AG 30 E.ON Ruhrgas International AG 71 SIDARSTEEL N.V. 112 VNG Verbundnetz Gas Verwaltungs- und Bet.-GmbH 31 E.ON Ruhrgas Thüga Holding GmbH 72 SIDMAR N.V. 113 VNG-Erdgascommerz GmbH 32 E.ON Sales & Trading GmbH 73 Staat Schweden 114 WEMAG AG 33 E.ON Sverige AB 74 Stadt Chemnitz 115 Weser-Ems-Energiebeteiligungen GmbH 34 E.ON Thüringer Energie AG 75 Stadt Darmstadt 116 Stadt Augsburg 35 E.ON Wasserkraft GmbH 76 Stadt Frankfurt am Main 117 Stadt Regensburg 36 E.ON Westfalen Weser AG 77 Stadt Kaiserslautern 37 EEG - Erdgas Erdöl GmbH 78 Stadt Landau 38 EEG - Erdgas Transport GmbH 79 Stadt Nürnberg 39 EnergieSüdwest AG 80 Stadt Zweibrücken 40 Erdgasversorgungsgesell. Thür.-Sa. mbH (EVG) 81 Stadtholding Landau in der Pfalz GmbH 41 EWE AG 82 Städtische Werke Nürnberg GmbH Network Figure 3: Giant 49 36 3
16
7
4 29 32 37 5 34 20 15 40 38 31 28 6 21 25 35 27 39 24 30
11 26 13 14 1 17 23
12 8
33
9 10
19
2 22
18 Giant 49 This is the Giant Component of the Complete Network where all links below 49% are deleted.
1 A. Friedr. Flender Aktiengesellschaft 21 Robert Bosch Stiftung GmbH 2 Automobiles Peugeot S.A. 22 S.I.P. Verwaltungsgesellschaft mbH 3 BBT Thermotechnik GmbH 23 SAS Autosystemtechnik Verwaltungs GmbH 4 Blaupunkt GmbH 24 Siemens Aktiengesellschaft 5 Bosch Rexroth Aktiengesellschaft 25 Siemens Aktiengesellschaft ﾖsterreich 6 BSH Bosch und Siemens Hausger 舩e GmbH 26 Siemens Beteiligungen Management GmbH 7 Buderus Aktiengesellschaft 27 Siemens Beteiligungsverwaltung GmbH & Co. OHG 8 Faurecia Automotive GmbH 28 Siemens Business Services Beteiligungs-GmbH 9 Faurecia Autositze GmbH & Co. KG 29 Siemens Business Services GmbH 10 Faurecia S.A. 30 Siemens Business Services GmbH & Co. OHG 11 Flender Holding GmbH 31 Siemens Real Estate GmbH & Co. OHG 12 Fujitsu Ltd. 32 Siemens Real Estate Management GmbH 13 Fujitsu Siemens Computers (Holding) B.V. 33 Sommer Allibert S.A. 14 Fujitsu Siemens Computers GmbH 34 Stadt Friedrichshafen 15 Kabel- und Drahtwerke Aktiengesellschaft 35 VVK Vers.-Verm.- und Verkehrs-Kontor GmbH 16 Luftschiffbau Zeppelin GmbH 36 ZEPPELIN GmbH 17 Osram GmbH 37 Zeppelin-Stiftung 18 PEUGEOT DEUTSCHLAND GMBH 38 ZF FRIEDRICHSHAFEN Aktiengesellschaft 19 Peugeot S.A. 39 ZF Getriebe GmbH 20 Robert Bosch GmbH 40 ZF Lenksysteme GmbH Network Figure 4: Giant 74
21 5 This is the Giant Component of the Complete 9 19 Network where all links below 74% are deleted.
26 23 1 Aldi GmbH & Co. KG Adelsdorf 2 Aldi GmbH & Co. KG Altenstadt 24 3 Aldi GmbH & Co. KG Bingen am Rhein 17 4 Aldi GmbH & Co. KG Bous 5 Aldi GmbH & Co. KG Donaueschingen 6 Aldi GmbH & Co. KG Ebersberg 4 2 7 Aldi GmbH & Co. KG Eichenau 8 Aldi GmbH & Co. KG Geisenfeld 9 Aldi GmbH & Co. KG Helmstadt 10 Aldi GmbH & Co. KG Kerpen 10 11 Aldi GmbH & Co. KG Ketsch 13 12 Aldi GmbH & Co. KG Kirchheim an d. Weinstr. 13 Aldi GmbH & Co. KG Langenfeld L. (Rheinland) 14 Aldi GmbH & Co. KG Langenselbold 15 Aldi GmbH & Co. KG Mahlberg 12 25 16 Aldi GmbH & Co. KG Mönchengladbach 28 17 Aldi GmbH & Co. KG Montabaur 18 Aldi GmbH & Co. KG Mörfelden-Walldorf 19 Aldi GmbH & Co. KG Mühlheim an der Ruhr 6 20 Aldi GmbH & Co. KG Murr 15 21 Aldi GmbH & Co. KG Rastatt 22 Aldi GmbH & Co. KG Rheinberg 23 Aldi GmbH & Co. KG Sankt Augustin 24 Aldi GmbH & Co. KG Wittlich 22 25 Aldi GmbH & Co. KG Aichtal 1 26 ALDI GmbH & Co. KG Regenstauf 27 Aldi GmbH & Co. KG Roth 11 28 Siepmann Stiftung 14
8 20
7 16 18 3 27 124 107 119 Network Figure 5: ALDI 81 85 71 105 102 74
127 32 106 28 12 58 103 95 79 91 60 26 53 47 110 67 125 90 50 35 8 37 126 78 129 14 136 68 111 54 25 36 112 80 39 116 59 46 86 64 6 122 10 7 3 18 82 55 45 19 83 63 24 51 88 42 9 134 117 40 56 16 33 120 13 17 2 61 73 62 118 15 21 115 27 101 66 87 92 84 20 57 11 65 75 89 96 128 113 114 76 30 109 44 4 70 108 77 43 97 38 34 135 5 22 23 52 41 31 99 48 29 133 130 69 104 49 72
132 123 94 121 98 1 93 100 131 ALDI 1 A. Dold GmbH 47 Aldi GmbH & Co. KG Radevormwald 93 Hahn GmbH 2 Albers GmbH 48 Aldi GmbH & Co. KG Rastatt 94 Hake GmbH 3 Aldi Einkauf GmbH & Co. oHG Essen 49 Aldi GmbH & Co. KG Rheinberg 95 Heckl GmbH 4 Aldi Einkauf GmbH & Co. oHG Mühlheim an der Ruhr 50 Aldi GmbH & Co. KG Rinteln 96 Heußinger GmbH 5 Aldi Einkauf GmbH Duisburg 51 Aldi GmbH & Co. KG Salzgitter 97 Hirtz GmbH 6 Aldi Einkauf GmbH Herten 52 Aldi GmbH & Co. KG Sankt Augustin 98 Hoffmann Beteiligungsgesellschaft mbH 7 ALDI Gesellschaft & Co. KG Großbeeren 53 ALDI GmbH & Co. KG Scharbeutz 99 Holger Schmidt GmbH 8 ALDI GmbH & Co. Beucha KG 54 Aldi GmbH & Co. KG Schloß Holte-Stukenbrock 100 Holger Schneider GmbH 9 Aldi GmbH & Co. KG Adelsdorf 55 ALDI GmbH & Co. KG Schwelm 101 Iders GmbH 10 Aldi GmbH & Co. KG Altenstadt 56 ALDI GmbH & Co. KG Seefeld 102 Jakobus-Stiftung 11 Aldi GmbH & Co. KG Bad Laasphe 57 ALDI GmbH & Co. KG Seevetal 103 Karl Albrecht 12 Aldi GmbH & Co. KG Bargteheide 58 ALDI GmbH & Co. KG Werl 104 Kehl GmbH 13 ALDI GmbH & Co. KG Berlin 59 ALDI GmbH & Co. KG Weyhe 105 Kenzler GmbH 14 ALDI GmbH & Co. KG Beverstedt 60 ALDI GmbH & Co. KG Wilsdruff 106 Kießl GmbH 15 Aldi GmbH & Co. KG Bingen am Rhein 61 Aldi GmbH & Co. KG Wittlich 107 Langenstroeher GmbH 16 Aldi GmbH & Co. KG Bous 62 ALDI GmbH & Co. KG Wittstock 108 Larberg GmbH 17 ALDI GmbH & Co. KG Datteln 63 Aldi GmbH & Co. Kommanditgesellschaft Aichtal 109 Lessner GmbH 18 Aldi GmbH & Co. KG Donaueschingen 64 Aldi GmbH & Co. Kommanditgesellschaft Greven 110 Liebisch GmbH 19 Aldi GmbH & Co. KG Ebersberg 65 ALDI GmbH & Co. Kommanditgesellschaft Regenstauf 111 Lukas Stiftung 20 Aldi GmbH & Co. KG Eichenau 66 Aldi GmbH & Co. Kommanditgesellschaft Roth 112 Markhoff GmbH 21 ALDI GmbH & Co. KG Essen 67 Aldi GmbH u. Co. KG Notdorf 113 Markus Kaffee GmbH 22 Aldi GmbH & Co. KG Geisenfeld 68 Berger GmbH 114 Markus Kaffee GmbH & Co. KG Herten 23 Aldi GmbH & Co. KG Helmstadt 69 Berthold Albrecht 115 Markus Kaffee GmbH & Co. KG Weyhe 24 ALDI GmbH & Co. KG Herten 70 Billen GmbH 116 Markus Stiftung 25 Aldi GmbH & Co. KG Hesel 71 Brehm GmbH 117 Michalek GmbH 26 Aldi GmbH & Co. KG Horst 72 Bröker GmbH 118 Müller GmbH 27 ALDI GmbH & Co. KG Hoyerswerder 73 Burgard GmbH 119 Neubold GmbH 28 ALDI GmbH & Co. KG Jarmen 74 Buttkus GmbH 120 Noack GmbH 29 Aldi GmbH & Co. KG Kerpen 75 Carolus-Stiftung 121 Otte GmbH 30 Aldi GmbH & Co. KG Ketsch 76 Daniel GmbH 122 Penkert GmbH 31 Aldi GmbH & Co. KG Kirchheim an der Weinstraße 77 David GmbH 123 Polossek GmbH 32 ALDI GmbH & Co. KG Könnern 78 Delschen GmbH 124 Reitzig GmbH 33 Aldi GmbH & Co. KG Langenfeld Langenfeld (Rheinland) 79 Diekhaus GmbH 125 Robinson GmbH 34 Aldi GmbH & Co. KG Langenselbold 80 Drees GmbH 126 Roettgen GmbH 35 ALDI GmbH & Co. KG Langenwetzendorf 81 Ebel GmbH 127 Sander GmbH 36 Aldi GmbH & Co. KG Lehrte 82 Eck GmbH 128 Siepmann Stiftung 37 ALDI GmbH & Co. KG Lingen (Ems) 83 Eden GmbH 129 Steinbrenner GmbH 38 Aldi GmbH & Co. KG Mahlberg 84 Ekrot GmbH 130 Theo Albrecht jun. 39 ALDI GmbH & Co. KG Meitzendorf 85 Elsner GmbH 131 Thull GmbH 40 ALDI GmbH & Co. KG Mittenwalde 86 Familie Albrecht 132 Thunig GmbH 41 Aldi GmbH & Co. KG Mönchengladbach 87 Fenten Gesellschaft mit beschränkter Haftung 133 Tölle GmbH 42 Aldi GmbH & Co. KG Montabaur 88 Feucht GmbH 134 Vollmer GmbH 43 Aldi GmbH & Co. KG Mörfelden-Walldorf 89 Frank Schröder GmbH 135 Weiland GmbH 44 Aldi GmbH & Co. KG Mühlheim an der Ruhr 90 Gerdes GmbH 136 Oertel-Stiftung 45 Aldi GmbH & Co. KG Murr 91 Goetsch GmbH 46 ALDI GmbH & Co. KG Nohra 92 Günther GmbH Network Figure 6: AMB Generali 110 113 43 81 53 73 109 82 51 50 56 24 44 49 25 65 115 14 3 17 15 2966 52 27 54 111 39 26 1 59 57 55 8 45 6 112 5 114 2 58 37 30 32 80 4 19
62 11 33 78 23 79 71 76 38 48 61 77 31 69 20 75
100 68 72 7 60 101 36 70 67 64 22 41 74 47 12 63 13 21 9 46 18 10 16 42 34 98 28 92 40 99 93 35 83 86 85 89 102 84 87 91 96 104 94 103 105 88 108 95 90 107 106 97 AMB Generali 1 Allianz Aktiengesellschaft 44 Hamburg-Mannheimer Sachversicherungs-AG 87 Finadin - S.p.A. Finanziaria di Investimenti 2 Allianz Deutschland AG 45 Hamburg-Mannheimer Versicherungs-Aktiengesellschaft 88 Compagnia Fiduciaria Nazionale S.p.A. 3 Allianz Finanzbeteiligungs GmbH 46 Ina Vita S.p.A. 89 Banca del Gottardo S.A. 4 Allianz Lebensversicherungs-Aktiengesellschaft 47 Italcementi Fabbriche Riunite Cemento S.p.A. 90 Sinergia Terza S.p.A. 5 Allianz Subalpina Società di assicurazioni e riassicurazioni 48 Italmobiliare S.p.A. 91 Canoe Securities S.A. 6 Asopos Vermögensverwaltungsgesellschaft 49 KARSTADT QUELLE Aktiengesellschaft 92 Hike Securities S.A. 7 Assicurazioni Generali S.p.A. 50 KARSTADT QUELLE Kunden-Service GmbH 93 Limbo Invest S.A. 8 AVIVA Plc. 51 KARSTADT QUELLE Service GmbH 94 Immobiliare Costruzioni IM.CO. S.p.A. 9 B and B Investissement S.C. Immobilière 52 KarstadtQuelle Lebensversicherung Aktiengesellschaft 95 SAIFIN - SAI Finanziaria S.p.A. 10 Banca di Roma S.p.A. 53 Leo Herl 96 Schweizerische LV. - und Rentenanstalt 11 Banca d'Italia S.p.A. 54 Madeleine Schickedanz 97 Giulia Maria Ligresti 12 Banca Mediosim Banca della Rete S.p.A. 55 Madeleine Schickedanz Vermögensverwaltungs B. GmbH 98 Jonella Ligresti 13 Banco di Sicilia S.p.A. 56 Madeleine Schickedanz Vermögensverw. GmbH & Co. KG 99 Gioacchino Paolo Ligresti 14 Barclays PLC 57 Martin Dedi 100 AMB Generali Holding AG 15 Bayerische Hypo- und Vereinsbank Aktiengesellschaft 58 Martin Dedi Vermögensverwaltungs Beteiligungs GmbH 101 Generali Beteiligungs-GmbH 16 Caisse Centrale des Assurances Mutuelles A.S.M. 59 Martin Dedi Vermögensverwaltungs GmbH & Co. KG 102 F. Mandori 17 Capital Research & Management Company 60 MEDIOBANCA Banca di Credito Finanziario S.p.A. 103 M. Ardesi 18 Capitalia S.p.A. 61 Merrill Lynch Investments Managers Group Ltd. 104 Familie Schickedanz 19 Carimonte Holding S.p.A. 62 Münchener Rückversicherungs-Gesellschaft AG 105 Putnam Investments. LLC 20 COMMERZBANK Aktiengesellschaft 63 Novara Vita S.p.A. 106 Soc. Reale Mutua di Assicurazioni 21 Compagnia di Assicurazione di Milano S.p.A. 64 Po Vita Compagnia di Assicurazioni S.p.A. 107 C. Gestioni 22 Compass S.p.A. 65 RAS Riunione Adriatica di Sicurtà S.p.A. 108 A. Spaggiari 23 Consortium S.r.l. 66 RB Vita S.p.A. 109 Crédit Industriel d'Alsace et de Lorraine S.A. 24 D.A.S. Deutscher Automobil Schutz Allgemeine R.-V. AG 67 Republik Frankreich 110 Crédit Industriel et Commercial (CIC) 25 D.A.S. Deutscher Automobil Schutz Versicherungs-AG 68 Sade Finanziaria S.p.A. 111 Legal & General Group PLC 26 DKV Deutsche Krankenversicherung Aktiengesellschaft 69 SIAT - Società Italiana Assicurazioni e Riassicurazioni - p.A. 112 Fidelity Investments Ltd. 27 DRESDNER BANK Aktiengesellschaft 70 Società per Amministrazioni Fiduciarie SPAFID S.p.A. 113 ERGO Achte Beteiligungsgesellschaft mbH 28 EFFE Finanziaria S.p.A. 71 Società per la Bonifica dei Terreni F. e per le I. A. - S.p.A. 114 MR ERGO Beteiligungen GmbH 29 ERGO Versicherungsgruppe Aktiengesellschaft 72 Société der Participation Financière Italmobiliare S.A. 115 KarstadtQuelle Finanz Service GmbH 30 Europäische Reiseversicherung Aktiengesellschaft 73 The Capital Group Companies Inc. 31 Fidelity International Limited 74 The Lawrence Re Ireland Ltd. 32 Fidelity Investments International 75 Tradinglab Banca S.p.A. 33 Fidelity Investments International Limited 76 Unicredit Banca d'Impresa S.p.A. 34 Financière du Perguet S.A.S. 77 Unicredit Banca Mobiliare S.p.A. 35 FinecoGroup S.p.A. 78 Unicredit Banca S.p.A. 36 FINSAI INTERNATIONAL S.A. 79 Unicredit Private Banking S.p.A. 37 Fondazione Cassa di Risparmio di Torino 80 UniCredito Italiano S.p.A. 38 Fondazione Cassa di Risparmio Verona. Vicenza. B. e A. 81 VICTORIA Lebensversicherung Aktiengesellschaft 39 Fondazione Cassamarca -Cassa di Risparmio della M. T. 82 VICTORIA Versicherung Aktiengesellschaft 40 Fondiaria - SAI S.p.A. 83 Roma Vita S.p.A. 41 GAN S.A. 84 Toro Assicurazioni S.p.A. 42 Generali Vita S.p.A. 85 DE AGOSTINI S.p.A. 43 Grisfonta AG 86 Premafin Finanziaria - S.p.A. Holding de Partecipazioni Network Figure 7: AXA 1 Allianz Aktiengesellschaft 2 Assurances Générales de France S.A. 17 18 3 AXA ASSURANCES VIE MUTUELLES 4 AXA Konzern Aktiengesellschaft 5 AXA S.A. 6 BNP Paribas S.A. 7 C.D.C. Caisse des Dépôts et Consignations 8 FINAXA SA 8 9 Kölnische Verwaltungs-AG 10 Les Ateliers de Construction du Nord de la France S.A. 6 11 Münchener Rückversicherungs-Gesellschaft AG 3 12 Republik Frankreich 13 Vinci B.V. 14 Crédit Agricole S.A. 15 Rue la Boétie SAS 9 5 16 Crédit Agricole Transactions SNC 20 17 AXA Assurances IARD Mutuelle S.A. 18 AXA Courtage Assurance Mutuelle 19 EURAZEO SA 20 Société Civile Haussmann Percier 10
4 16
19 13
14 15
2
12 7 1 11 Network Figure 8: BMW
1 Allianz AG 10 14 13 2 Allianz Deutschland AG 3 Allianz Finanzbeteiligungs GmbH 4 Allianz Lebensversicherungs-AG 8 5 Asopos Vermögensverwaltungsgesellschaft 6 Bayerische Motoren Werke AG 7 DRESDNER BANK AG 8 Johanna Quandt 9 Johanna Quandt GmbH & Co. KG für Automobilwerte 10 Münchener Rückversicherungs-Gesellschaft AG 11 Stefan Quandt 12 Stefan Quandt GmbH & Co. KG für Automobilwerte 1 13 Susanne Klatten 6 9 14 Susanne Klatten GmbH & Co. KG für Automobilwerte
12
2
11 3 7
4
5 Network Figure 9: Commerzbank
95 96 85 94 89 97 90 102 100 84 104 18 91 101 88 75 87 83 34 13 103 99 93 86 35 9 40 10 63 105 12 92 23 21 28 42 98 36 46 106 70 16 60 7 69 68 61 41 22 76 79 74 31 64 67 72 38 20 78 11 71 48 77 80 33 58 37 47 107 62 39 66 32 8 59 30 57 112 19 15 110 56 26 5 3 44 65 29 25 55 49 17 109 27 1 45 14 51 54 81 113 52 73 50 24 111 82 108 53 43 6 4 2 Commerzbank 1 Allianz AG 44 Hamburg-Mannheimer Sachversicherungs-AG 87 Finadin - S.p.A. Finanziaria di Investimenti 2 Allianz Deutschland AG 45 Hamburg-Mannheimer Versicherungs-AG 88 Compagnia Fiduciaria Nazionale S.p.A. 3 Allianz Finanzbeteiligungs GmbH 46 Ina Vita S.p.A. 89 Banca del Gottardo S.A. 4 Allianz Lebensversicherungs-AG 47 Italcementi Fabbriche Riunite Cemento S.p.A. 90 Sinergia Terza S.p.A. 5 Allianz Subalpina Società di assicurazioni e riassicurazioni 48 Italmobiliare S.p.A. 91 Canoe Securities S.A. 6 Asopos Vermögensverwaltungsgesellschaft 49 KARSTADT QUELLE AG 92 Hike Securities S.A. 7 Assicurazioni Generali S.p.A. 50 KARSTADT QUELLE Kunden-Service GmbH 93 Limbo Invest S.A. 8 AVIVA Plc. 51 KARSTADT QUELLE Service GmbH 94 Immobiliare Costruzioni IM.CO. S.p.A. 9 B and B Investissement S.C. Immobilière 52 KarstadtQuelle Lebensversicherung AG 95 SAIFIN - SAI Finanziaria S.p.A. 10 Banca di Roma S.p.A. 53 Leo Herl 96 Schweizerische LV- und Rentenanstalt 11 Banca d'Italia S.p.A. 54 Madeleine Schickedanz 97 Giulia Maria Ligresti 12 Banca Mediosim Banca della Rete S.p.A. 55 Madeleine Schickedanz Vermögensverw. Beteil. GmbH 98 Jonella Ligresti 13 Banco di Sicilia S.p.A. 56 Madeleine Schickedanz Vermögensverw. GmbH & Co. KG 99 Gioacchino Paolo Ligresti 14 Barclays PLC 57 Martin Dedi 100 F. Mandori 15 Bayerische Hypo- und Vereinsbank AG 58 Martin Dedi Vermögensverwaltungs Beteiligungs GmbH 101 M. Ardesi 16 Caisse Centrale des Assurances Mutuelles A. S. M. 59 Martin Dedi Vermögensverwaltungs GmbH & Co. KG 102 Familie Schickedanz 17 Capital Research & Management Company 60 MEDIOBANCA Banca di Credito Finanziario S.p.A. 103 Putnam Investments. LLC 18 Capitalia S.p.A. 61 Merrill Lynch Investments Managers Group Ltd. 104 Soc. Reale Mutua di Assicurazioni 19 Carimonte Holding S.p.A. 62 Münchener Rückversicherungs-Gesellschaft AG 105 C. Gestioni 20 COMMERZBANK AG 63 Novara Vita S.p.A. 106 A. Spaggiari 21 Compagnia di Assicurazione di Milano S.p.A. 64 Po Vita Compagnia di Assicurazioni S.p.A. 107 Crédit Industriel d'Alsace et de Lorraine S.A. 22 Compass S.p.A. 65 RAS Riunione Adriatica di Sicurtà S.p.A. 108 Crédit Industriel et Commercial (CIC) 23 Consortium S.r.l. 66 RB Vita S.p.A. 109 Legal & General Group PLC 24 D.A.S. Deutscher Automobil Schutz Allgemeine RV-AG 67 Republik Frankreich 110 Fidelity Investments Ltd. 25 D.A.S. Deutscher Automobil Schutz Versicherungs-AG 68 Sade Finanziaria S.p.A. 111 ERGO Achte Beteiligungsgesellschaft mbH 26 DKV Deutsche Krankenversicherung AG 69 SIAT - Società Italiana Assicurazioni e Riassicurazioni - p.A. 112 MR ERGO Beteiligungen GmbH 27 DRESDNER BANK AG 70 Società per Amministrazioni Fiduciarie SPAFID S.p.A. 113 KarstadtQuelle Finanz Service GmbH 28 EFFE Finanziaria S.p.A. 71 Società per la Bonifica dei Terreni F. e per le I. A. - S.p.A. 29 ERGO Versicherungsgruppe AG 72 Société der Participation Financière Italmobiliare S.A. 30 Europäische Reiseversicherung AG 73 The Capital Group Companies Inc. 31 Fidelity International Limited 74 The Lawrence Re Ireland Ltd. 32 Fidelity Investments International 75 Tradinglab Banca S.p.A. 33 Fidelity Investments International Limited 76 Unicredit Banca d'Impresa S.p.A. 34 Financière du Perguet S.A.S. 77 Unicredit Banca Mobiliare S.p.A. 35 FinecoGroup S.p.A. 78 Unicredit Banca S.p.A. 36 FINSAI INTERNATIONAL S.A. 79 Unicredit Private Banking S.p.A. 37 Fondazione Cassa di Risparmio di Torino 80 UniCredito Italiano S.p.A. 38 Fondazione Cassa di Risparmio Verona. Vicenza. B. e A. 81 VICTORIA Lebensversicherung AG 39 Fondazione Cassamarca -Cassa di Risparmio della M. T. 82 VICTORIA Versicherung AG 40 Fondiaria - SAI S.p.A. 83 Roma Vita S.p.A. 41 GAN S.A. 84 Toro Assicurazioni S.p.A. 42 Generali Vita S.p.A. 85 DE AGOSTINI S.p.A. 43 Grisfonta AG 86 Premafin Finanziaria - S.p.A. Holding de Partecipazioni Network Figure 10: DaimlerChrysler
1 DaimlerChrysler AG 7 2 DB Value GmbH 3 Deutsche Bank AG 4 Dubai Holding Ltd. 5 Dubai International Capital Ltd. 6 Emirat Kuwait 7 Mohammed bin Rashid AL Maktoum
4
5
1
2
3 6 Network Figure 11: Ergo 19 8 46 18 47
48 7
42 43 12 13 21 24 16 14 20 35 50 41 22 27 34 9 29 49 11 25 28 10 39 36 3 37 17 5 40
33 26 38
15 1
32 44 31 2 6 45 23 30 4 Ergo 1 Allianz AG 26 KARSTADT QUELLE AG 2 Allianz Deutschland AG 27 KARSTADT QUELLE Kunden-Service GmbH 3 Allianz Finanzbeteiligungs GmbH 28 KARSTADT QUELLE Service GmbH 4 Allianz Lebensversicherungs-AG 29 KarstadtQuelle Lebensversicherung AG 5 Allianz Subalpina Società di assicurazioni e riassicurazioni 30 Leo Herl 6 Asopos Vermögensverwaltungsgesellschaft 31 Madeleine Schickedanz 7 AVIVA Plc. 32 Madeleine Schickedanz Vermögensverwaltungs Beteiligungs GmbH 8 Barclays PLC 33 Madeleine Schickedanz Vermögensverwaltungs GmbH & Co. KG 9 Bayerische Hypo- und Vereinsbank AG 34 Martin Dedi 10 Capital Research & Management Company 35 Martin Dedi Vermögensverwaltungs Beteiligungs GmbH 11 Carimonte Holding S.p.A. 36 Martin Dedi Vermögensverwaltungs GmbH & Co. KG 12 D.A.S. Deutscher Automobil Schutz Allgemeine Rechtsschutz-Versicherungs-AG37 Münchener Rückversicherungs-Gesellschaft AG in München 13 D.A.S. Deutscher Automobil Schutz Versicherungs-AG 38 RAS Riunione Adriatica di Sicurtà S.p.A. 14 DKV Deutsche Krankenversicherung AG 39 RB Vita S.p.A. 15 DRESDNER BANK AG 40 The Capital Group Companies Inc. 16 ERGO Versicherungsgruppe AG 41 UniCredito Italiano S.p.A. 17 Europäische Reiseversicherung AG 42 VICTORIA Lebensversicherung AG 18 Fidelity Investments International 43 VICTORIA Versicherung AG 19 Fidelity Investments International Limited 44 Crédit Industriel d'Alsace et de Lorraine S.A. 20 Fondazione Cassa di Risparmio di Torino 45 Crédit Industriel et Commercial (CIC) 21 Fondazione Cassa di Risparmio Verona, Vicenza, Belluno e Ancona 46 Legal & General Group PLC 22 Fondazione Cassamarca -Cassa di Risparmio della Marca Trivigniana 47 Fidelity Investments Ltd. 23 Grisfonta AG 48 ERGO Achte Beteiligungsgesellschaft mbH 24 Hamburg-Mannheimer Sachversicherungs-AG 49 MR ERGO Beteiligungen GmbH 25 Hamburg-Mannheimer Versicherungs-AG 50 KarstadtQuelle Finanz Service GmbH Network Figure 12: Metro
17 1 1. HSB Beteiligungsverw. GmbH & Co. KG 18 2 1. HSB Verwaltung GmbH 3 Beisheim Holding GmbH 4 BVG Beteiligungs- und Verm.verw.GmbH 5 Dr. Michael Schmidt-Ruthenbeck 6 Familie Haniel 7 Franz Haniel & Cie. GmbH 2 8 Gebr. Schmidt GmbH & Co. KG 16 3 9 Haniel Finance B.V. 10 Haniel Finance Deutschland GmbH 11 METRO AG 12 Metro Vermögensverw. GmbH 15 13 Metro Vermögensverw. GmbH & Co. KG 1 14 O.B. Betriebs GmbH 15 O.B.V. Vermögensverw. mbH 16 O.B.V. Vermögensverw. mbH & Co. KG 17 Prof. Dr. Otto Beisheim 14 18 Prof. Otto Beisheim-Stiftung 19 SUPRA Holding AG 20 Supra Trust 11 13 21 Suprapart AG 21
4 12
10 19 9 8
7
20 5
6 Network Figure 13: Deutsche Post
1 Bundesländer 1 2 2 Bundesrepublik Deutschland 3 Deutsche Post AG 4 KfW Bankengruppe
4
3 Network Figure 14: Deutsche Telekom
1 Blackstone Group L.P. 2 Bundesländer 3 3 Bundesrepublik Deutschland 4 Deutsche Telekom AG 5 KfW Bankengruppe
4 5
1 2 Network Figure 15: Volkswagen 15
18 1
12 1 Brandes Investment Partners Inc. 2 Dr. Ferdinand Piëch 3 Dr. Hans-Michel Piëch 4 Dr. Ing. h.c. F. Porsche Aktiengesellschaft 5 Dr. Wolfgang Porsche 2 19 6 Familie Porsche 7 Familie Porsche Beteiligung GmbH 8 Familien Porsche-Daxer-Piëch Beteiligung GmbH 9 Ferdinand Alexander Porsche 10 Ferdinand Piëch GmbH 11 Gerhard Porsche 10 8 12 Hannoversche Beteiligungsgesellschaft mbH 13 Hans-Michel Piëch GmbH 14 Hans-Peter Porsche 15 Land Niedersachsen 16 Louise Daxer-Piëch 17 Porsche GmbH 18 The Capital Group Companies Inc. 4 19 Volkswagen Aktiengesellschaft
13
3
9
17 7
16
6 5
14 11